Semantics of the northern khanty salient article: definiteness, salience, and obviation

Full Text

1. Introduction

Many Uralic languages make use of possessive agreement markers (possessives) that index the person and number of the Possessor on the possessed noun phrase (NP). Such markers are well-known to exhibit non-possessive, “extended” or “discourse” uses which resemble the definite articles of European languages to a high extent [Fraurud 2001; Kuznetsova 2003; Nikolaeva 2003]. Recently, several authors have argued that the possessives they investigate have in fact become grammaticalized as independent markers with determiner-like functions [É. Kiss 2018; É. Kiss, Tánczos 2018; Halm 2018; Serdobolskaya et al. 2019]. However, almost no detailed semantic analyses of such “unpossessive markers” have been provided so far.¹

One example of such grammaticalization of an unpossessive determiner from a possessive marker is provided by the second-person singular Possessive in the Kazym dialect of Northern Khanty, which was first described in [Mikhailov 2021a]. In the following examples, it is used with a topical referent (1) and with a demonstrated referent (2).

(1) [“I was walking along the street when I saw a dog.”]²

amp-en ma pɛλ-am-a χurət-ti pit-əs

dog-poss.2sg I at-poss.1sg-dat bark-nfin.npst become-pst[3sg]

‘The dog started barking at me.’

(2) [A mother is walking with her child. The child points at a flower:]

ma tum lipt-en mɛnəm-ti λaŋχa-λ-əm

I that flower-poss.2sg pick-nfin.npst want-npst-1sg

‘I want to pick that flower.’

In both examples there does not seem to be any discourse-internal relation between the addressee and the referent of the highlighted noun phrase. Intuitively, this suggests that we are not dealing with the Proper 2sg Possessive³ here, but rather with some Unpossessive. (I will substantiate this claim with empirical arguments below.) I call this Unpossessive the “Salient Article”, for reasons to become clear in the discussion to follow.⁴

The goals of this paper are three-fold. Firstly, to provide a detailed empirical study of a single Uralic unpossessive, namely, the Salient Article poss.2sg of Kazym Khanty. Secondly, to give a motivated semantic analysis that correctly reflects the Salient Article’s conditions of use. And lastly, to situate the Salient Article in the typology of definiteness markers (e.g., [Schwarz 2019]).

I will argue for the following claims:

1. Noun phrases with the Salient Article are definite.
2. The Salient Article does not directly encode syntactic role or information-structural distinctions.
3. The Salient Article marks the most salient discourse referent in the given context.

Thus, I argue that the Kazym Khanty Salient Article instantiates a hitherto unattested definite determiner type. Even though it is similar to other topic-related phenomena in the world’s languages such as the Obdorsk Khanty Agreeing Objects [Nikolaeva 2001] or the Blackfoot Proximate Marking [Bliss 2017], the Kazym Khanty Salient Article does not directly encode either topichood (contra [Mikhailov 2021a]) or proximate status (contra [Muravyev 2022a]) but is used exclusively as a definite determiner.

The rest of the paper is structured as follows. In Section 2, I introduce the Kazym dialect of Northern Khanty, give the basics of the Kazym Khanty possessive agreement system, and argue for the independence of the Salient Article from the Proper poss.2sg. Section 3 introduces the theoretical notions to be employed in what follows, alongside cross-linguistic “standards of comparison” for our description of the Salient Article. In Section 4, I describe the Salient Article in detail and argue against a syntactic account of its distribution. In Section 5, I propose and motivate a (semiformal) semantic analysis of the Salient Article, based on Roberts’ [2003] semantics for definite NPs.

This study thus contributes to the recent trend of theoretically and typologically informed inquiries into non-possessive uses of Uralic possessives.

2. Basic information about the Kazym dialect of Northern Khanty

2.1. Language and methodology

Northern Khanty is an endangered Uralic language of the Khantyic branch⁵ spoken by the Ob’ river and its tributaries (Kazym, Kunovat, Synja, etc.) by about 8865 speakers which makes 37 % of the ethnic Khanty according to the Russian Census of 2010. The Kazym dialect has more than 1700 speakers [Kaksin 2010] with the overwhelming majority older than 50 years old [Aristova 2023].

The data used in this study were collected over a period from 2018 to 2023 during the field trips to the Kazym village (Khanty-Mansi Autonomous Region−Yugra, Russia), led by Svetlana Toldova and Alexey Kozlov and organized collectively by the HSE University (Moscow) and the Lomonosov Moscow State University. I elicited data from up to thirteen speakers following Lisa Matthewson’s methodology of semantic fieldwork [Matthewson 2004]. Each example was judged by at least three speakers. The speakers were presented with Russian stimuli in a context and were asked to translate them to Khanty. In most cases the context was also translated. Then, using the Khanty translation, the speakers were requested to provide acceptability judgements for particular forms in said context. Sometimes the speakers’ comments were used to create further stimuli to test the hypotheses suggested by their comments.

2.2. Kazym dialect Possessives

Like most Uralic languages, the Kazym dialect has possessive agreement markers which index the person-number features of the Possessor NP in adnominal possessive noun phrases (3). Possessives are obligatorily used with pronominal Possessors and may be absent when Possessors are lexical, i.e., non-pronominal (see [Muravyev 2022a] and §3.3 below for a preliminary hypothesis about Possessives with lexical Possessors).

(3) năŋ kătˊ-en moś-λ

you.sg cat-poss.2sg purr-npst[3sg]

‘Your cat purrs.’

The paradigm is presented in Table 1 (p. 11) with 3×3 person-number combinations for the Possessor and three numbers (singular, dual, plural) of the possessee since possessee number triggers allomorphic alternations in the Possessives.

Table 1. Possessive suffixes of the Kazym dialect of Northern Khanty (data collected by the field project under S. Toldova and A. Kozlov)

Possessives may appear with or without an explicit Possessor, e.g., ‘your dog’ may be expressed both as năŋ amp-en [you.sg dog-poss.2sg] and as amp-en [dog-poss.2sg]. Possessives never attach to a noun already marked with a Possessive regardless of their function: thus, forms like *amp-əλ-en [dog-poss.3sg-poss.2sg] (intended: ‘this dog of his’ or ‘his dog that you like’, etc.) are barred.⁶ This entails that, if a noun is already marked with a Possessive, it cannot be marked with the Salient Article in virtue of this morphosyntactic restriction.

2.3. The independence of the Salient Article from the Proper poss.2sg

I assume following [Mikhailov 2021a] that the Salient Article is an independent entity, distinct from the Proper poss.2sg. This assumption is supported by a range of data showing that the Salient Article and the Proper poss.2sg behave differently.

Morphosyntactically, the Salient Article does not co-occur with a Possessor NP (4) and never varies in person and number, see (5) where the addressee is plural but this does not affect the choice of the marker.

(4) [“I was walking along the street when I saw a dog.”]

(#năŋ) amp-en ma pɛλ-am-a χurət-ti pit-əs

you.sg dog-poss.2sg I at-poss.1sg-dat bark-nfin.npst become-pst[3sg]

‘The dog started barking at me.’⁷

(5) [A mother is telling her children: “I was walking along the street when I saw a dog.”]

amp-en/#-ən ma pɛλ-am-a χurət-ti pit-əs

dog-poss.2sg/-poss.2nsg I at-poss.1sg-dat bark-nfin.npst become-pst[3sg]

‘The dog started barking at me.’

This suggests that, unlike Proper Possessives, the Salient Article neither selects for a Possessor NP, nor agrees with any such NP for person-number features, nor implies a Possessor referent semantically. I take this to be sufficient evidence to treat it as a separate marker with distinct semantic and morphosyntactic properties.

Example (6) similarly shows that the morpheme used with demonstratives does not vary in person-number features despite the presence of a plural addressee. I also take this to indicate that it is not a Possessive, but rather the Salient Article.

(6) [A mother asked her children to do the dishes. Later she takes a pot and finds that it is still dirty. She says:]

niŋ wan-t-an-ən tăm pʉt-en/#-ən śistam?

you.nsg see-nfin.npst-2nsg-loc this cup-poss.2sg/-poss.2nsg clean

‘Do you think this pot is clean ’

As I will argue in §4, the Salient Article is in fact a special kind of a definite determiner which is semantically quite far apart from Possessives.

Two other Kazym Khanty Unpossessives that will surface in the foregoing discussion are the Associative Possessive and the Proprial Article [Mikhailov 2023], also briefly introduced here.

The Associative Possessive is, roughly speaking, a definite counterpart to the Proper Possessive. This use of Northern Khanty possessives was first described in [Nikolaeva 2003].⁸ It is found with any person and number and indicates that the referent of the marked NP is presupposed to stand in some contextually given relation to another activated referent. Thus, in (7) the passport is presupposed to stand in a relation to the speaker in virtue of the speaker’s finding it in the street in the preceding context.

(7) [“I found somebody’s passport in the street. I went to the town administration. Met a friend there and talked to her for some time.”]

nɛm nɛpek-ɛm suwet-ən χăj-s-ɛm

name paper-poss.1sg council-loc leave-pst-1sg>sg

‘[Then] I left the passport at the administration. [Let them find the owner.]’

In the discussion to follow, we will see that sometimes when the Salient Article is unavailable, the Associative Possessive is used instead (see §4.3).

The Proprial Article is found with human names in argument positions as in (8) (cf. [Muñoz 2019]). It is used to derive referring expressions from human name predicates.

(8) wɵntər-en jʉχ šɵp sewr-əs

A.-poss.2sg wood piece cleave-pst[3sg]

‘Andrej cleaved a log.’

As [Mikhailov 2023] shows, the Proprial Article differs from the Salient Article among other things in that it derives rigid designators [Kripke 1980], i.e., NPs that cannot co-vary in reference with any quantifiers. The full presentation of the differences between the two markers is beyond the scope of this paper, so I will simply assume that the Proprial Article is independent in what follows.

To prepare for the discussion to come, in the next section I introduce the relevant theoretical concepts and cross-linguistic evidence that we will appeal to in testing different hypotheses about the Salient Article.

3. Basic theoretical notions

3.1. Definiteness and salience

3.1.1. Definiteness

Following Craige Roberts [2003] among many others (see [Heim 2019]), I understand definiteness to consist in two requirements: familiarity and uniqueness. To illustrate Roberts’ analysis let us consider the distribution of the English Definite Article.

Both uniqueness and familiarity are notions that have been proposed as the core of the English Definite Article’s semantics in the tradition prior to Roberts.⁹

Uniqueness requires there to be one and only one referent satisfying the descriptive content of the noun phrase (NP) in the model. Thus, in (9) the Definite Article is used because the Queen of England is a globally unique entity. In (10), upon hearing the sentence the addressee will be justified in assuming that there was one and only one button in the box. (For plural NPs, uniqueness is often referred to as maximality: plural definites require the referent to be the unique maximal set satisfying the descriptive content of the NP [Link 1983].)

(9) The Queen of England had a bad year in 1993. [Roberts 2003: 290]

(10) I found a box in my attic the other day. I opened the lid and pushed the button I found inside. You won’t believe what happened. (adapted from [Ibid.])

Familiarity requires there to be a discourse referent¹⁰ introduced in prior context that satisfies the definite NP’s descriptive content. According to Irene Heim’s seminal thesis [1982], the use of definite NPs (including personal pronouns) and indefinite NPs is governed by the Extended Novelty-Familiarity Condition ([Heim 1982: 396] cited after [Roberts 2003: 295 ff.]). Under this Condition, the use of a definite NP presupposes the existence of a corresponding discourse referent in the context (familiarity). The use of an indefinite NP presupposes the absence of a corresponding discourse referent, so that indefinite NPs introduce novel discourse referents (novelty).

Thus, in (11) the Definite Article is obligatorily used upon second mention of the glass since the context does provide an appropriate discourse referent, which has been introduced by the first mention with the Indefinite Article. On the other hand, the Indefinite Article is infelicitous in the second clause since that would violate its novelty presupposition, whereas the Definite Article cannot be used upon first mention of the glass since it is not familiar in any sense at that point.

(11) A / #the wine glass broke last night. The / #a glass had been very expensive. (based on [Roberts 2003: 293])

Note also that this example presents a problem for theories of uniqueness since the Definite Article here does not seem to require uniqueness of the glass in the world or even in the relevant household.

Roberts points to the fact that discourse referents are not only introduced by linguistic material (which is the case of strong familiarity / anaphoricity) but may also be introduced via entailment from the common ground or from the extralinguistic context of the speech situation. This observation is captured under her notion of weak familiarity:

(12) Weak Familiarity of Discourse Referents [Roberts 2003: 298, 304, 306]
A discourse referent is weakly familiar iff the existence of the entity in question is entailed by the interlocutors’ common ground.

Both uniqueness and familiarity effects may be observed in cases involving bridging [Clark 1975] such as (13). In (13a), the introduction of a unicycle into the context entails the existence of its wheel, which licenses the Definite Article since the wheel is familiar via entailment from the common ground. However, the Definite Article is barred in (13b) even though, presumably, the tires of a car also become familiar as soon as a car is introduced. The reason that this utterance is infelicitous is the fact that uniqueness is not satisfied here since a car has four tires unlike a unicycle which only has one wheel.

(13)

1. Every unicycle had a spoke missing from the wheel.
2. #Every car had a puncture in the tire. (adapted from [Roberts 2003: 291])

Roberts takes these (and other) examples to show that the Definite Article has both a (weak) familiarity and a uniqueness requirement, unifying the two analytical traditions.

But what about the glass example (11) above, in which uniqueness in the strict sense does not hold This problem dissolves if one adopts, following Roberts, informational uniqueness. Unlike semantic uniqueness, standardly assumed in works prior to Roberts, informational uniqueness does not require uniqueness in the world but is only concerned with uniqueness among weakly familiar discourse referents. Roberts’ analysis of definite NPs is summarized below.

(14) Informational Existence and Uniqueness of Definite NPs (informal) [Roberts 2003: 308]¹¹

Given a context C, use of a definite NP_i presupposes that it has as antecedent¹² a discourse referent x_i which is:

1. weakly familiar in C, and
2. unique among discourse referents in C in being contextually entailed to satisfy the descriptive content of NP_i.

Still, there are cases in which it seems that the conditions in (14) are not satisfied but the Definite Article is felicitously used. For instance, consider (15) below, repeated from (10). Clearly, neither weak familiarity nor informational uniqueness are satisfied in this context, as no familiar and unique button entity has been introduced in (or is entailed by) the prior discourse. What happens here, however, is that the addressee accommodates semantic uniqueness and weak familiarity (“there was one and only one button that the speaker found inside the box”) and this entails informational uniqueness. As Roberts puts it, “since the presupposition of informational uniqueness in definite NPs can be satisfied by semantic uniqueness plus weak familiarity, speakers actually use this fact and the expectation that a hearer will cooperatively accommodate (when retrievable and plausible [see below — S. M.]) to convey the information that a definite NP’s intended denotation is semantically unique” [Roberts 2003: 319].¹³

(15) I found a box in my attic the other day. I opened the lid and pushed the button I found inside. You won’t believe what happened.

This analysis crucially relies on the notion of presupposition accommodation [Lewis 1979], a mechanism whereby the addressee adds information to her beliefs in order to make sense of an apparent presupposition failure on the speaker’s part. Accommodation is commonly used by speakers to background newly introduced information. It is generally available given that the following conditions obtain:

(16) Necessary Conditions on Presupposition Accommodation [Roberts 2003: 302–303]

1. Retrievability: what the hearer is to accommodate is easily inferable, by virtue of its salience and relevance to the immediate context, and
2. Plausibility: the accommodated material is unobjectionable.

One last relevant observation concerns the infelicity of the Indefinite Article with superlative NPs, whose descriptive content guarantees semantic uniqueness (17). Under a presuppositional view of the uniqueness requirement [Frege 1892; Strawson 1950], the obligatory use of the Definite Article in such cases can be accounted for via Irene Heim’s Maximize presupposition! principle [Heim 1991], which, informally speaking, requires the Definite Article to be used in all cases where its uniqueness presupposition is satisfied in the context. The use of the Indefinite Article leads to an anti-uniqueness inference: the addressee infers that the speaker used an Indefinite Article where a Definite Article was possible because the latter’s requirements are not satisfied in the context — there is more than one referent satisfying the NP’s content. With superlatives, the anti-uniqueness inference gives rise to a contradiction, hence the infelicity in (17).

(17) Last weekend we climbed the / #a biggest mountain in West Virginia. (adapted from [Roberts 2003: 294])

This concludes our discussion of the English Definite Article, and we now turn to other types of definite determiners found cross-linguistically.

3.1.2. Anaphoric articles

In recent years starting with Florian Schwarz’s dissertation [2009], there has been a rise in cross-linguistic research on definite determiner semantics (for an overview and for references see [Schwarz 2019]).

Schwarz investigates the phenomenon of German preposition-determiner contraction and argues that the “Weak” (contracted) Articles correspond to uniqueness-based definites while the “Strong” (non-contracted) Articles correspond to anaphoricity-based definites (where anaphoricity roughly corresponds to Roberts’ strong familiarity¹⁴,¹⁵).

Following his work, Ruby Arkoh and Lisa Matthewson [2013] show that determiners like the German Strong Article are attested cross-linguistically drawing on data from Akan (< Kwa < Niger-Congo). Below, I illustrate this anaphoric article category type with data from Akan. (Note that Arkoh and Matthewson argue that the Akan Definite Article nʊ and the German Strong Article are semantically equivalent, so everything said below about the former applies to the latter. See also [Bombi 2018] for a differing view on the Akan Definite Article.)

Much like the in English, the Akan Definite Article is not available upon first mention of a novel referent such as èkùtú ‘orange’ in (18) and it is obligatory upon second mention where the referent is already familiar.

(18) [Beginning of conversation.]

Mʊ̀-tɔ́-ɔ̀ èkùtú (*nʊ́). Èkùtú *(nʊ́) yὲ dὲw pápá

1sg.subj-buy-past orange (*fam) orange *(fam) be nice good

‘I bought an orange. The orange was really tasty.’ [Arkoh, Matthewson 2013: 2]

Unlike the, however, the Akan Definite Article cannot be used with novel unique referents. Thus, in (19) neither nyímpá ‘person’ nor ɔ̀sɪ̀rán ‘moon’ can be marked with nʊ, despite both of them being unique.¹⁶ Cf. the English translation of (19) where omitting the in either case is impossible.

(19) Ámstrɔ̀ŋ nyɪ́ nyímpá áà ó-dzí-ì kán tú-ù kɔ́-ɔ̀ ɔ̀sɪ̀rán dʊ̀

1. is person rel 3sg.subj-eat-past first fly-past go-past moon top

‘Armstrong was the first person to fly to the moon.’ [Ibid.]

The Akan data show that a clear empirical difference exists between two types of definite determiners, so it makes sense to ask, when investigating a definite determiner, which type it resembles the most.¹⁷ As I will show in §4, the Northern Khanty Salient Article has an even narrower distribution that will be accounted for by appealing to salience to which we now turn.

3.1.3. The Salient Article of Bulu

Salience has been appealed to in the analysis of definites before. In particular, Roberts [2003: 330] argues that personal pronouns are essentially definites, presupposing (14), with the additional requirement that the antecedent be salient in the context and that it be the most salient discourse referent satisfying the descriptive content of the pronoun (i.e., person, number, and gender).

To illustrate this analysis, consider (20). In this example, two female referents are introduced. Using the definite NP the woman to refer to either of them is infelicitous because it would violate informational uniqueness: neither woman is unique among the discourse referents in satisfying the NP content. On the other hand, using the pronoun she is clearly felicitous, and the pronoun would refer to the latter of the two women since she is the most salient one due to her most recent mention. Thus, “[a]naphora resolution for pronouns effectively screens out non-salient portions of the discourse and the discourse referents therein, so that only salient discourse referents are accessible to serve as antecedents” [Ibid.: 323].¹⁸

(20) A woman entered from stage left. Another woman entered from stage right. She / #the woman was carrying a basket of flowers. (adapted from [Ibid.: 324])

Drawing on data from Bulu (< Bantu, Cameroon), Jefferson Barlew [2014] has argued that the notion of salience is relevant to the cross-linguistic study of definite determiner semantics.

Following Roberts’ later work [2011] (see also [Gundel et al. 1993; Grosz et al. 1995]), Barlew takes salience to be a function of attention capture [2014: 623–624]. A referent that is salient to a given individual is a referent that she is paying attention to. Just like attention, salience is a gradient notion so that a referent may be salient to a higher or lesser degree.

Assuming that information sharing is a goal of the interlocutors [Grosz et al. 1995; Roberts 2011; 2022], the interlocutors will pay attention to discourse referents that are relevant to addressing the current question(s) under discussion (QUD).¹⁹ Roughly speaking, questions under discussion are questions that need to be addressed by the interlocutors in order to share information on the current topic under discussion or to achieve their other relevant goals.²⁰

Finally, it must be noted that for determining salience only the addressee’s attentional state is relevant [Prince 1992; Gundel et al. 1993]. The speaker’s attentional state may be assumed, since the speaker necessarily pays attention to any discourse referent that she refers to. Otherwise, it would be impossible to refer to it.

Barlew shows that the Salient Article of Bulu tè is restricted precisely to discourse referents that the addressee attends to. Thus, in (21a) where the Salient Article is used, the referent of kálàtà tè ‘the book’ is important to both speaker and addressee, and this ensures that it is attended to in the context involving a robbery. (The context suggests an implicit QUD “Is the book safe ” which, based on the discussion above, guarantees the discourse referent’s salience.) In (21b), on the other hand, the addressee is uninterested in the book even though we may assume that the book is familiar to him, and the Salient Article is infelicitous since the addressee cannot be assumed to attend to the discourse referent.

(21)

1. [Andung and Abondo have a special book that is a family heirloom passed down from Andung’s mother. It has family genealogy written inside it. They always keep it on the nightstand beside their bed. One day when they come home, they find their house has been broken into. When they come to the nightstand, they see that the book is gone. They exchange a glance, and then Andung says:]
   kálàtà tè à nə̀ ndʒáŋán
   book te pn₁ cop missing
   ‘The book is missing.’ [Barlew 2014: 626]
2. [Minimally different from (a), except that Andung is speaking to her teenage son, who does not care about genealogy or family heirlooms.]
   #kálàtà tè à nə̀ ndʒáŋán
   book te pn₁ cop missing
   Intended: ‘The book is missing.’ [Ibid.]

Barlew further argues that the Salient Article does not require maximal salience, as personal pronouns do on Roberts’ account, but rather requires any (positive) level of salience with informational uniqueness. His argument is based on examples modelled after the English example (20). The Salient Article, as it turns out, is infelicitous in this kind of a context much like the English Definite Article and unlike personal pronouns [Barlew 2014: 630–631]. According to Barlew, this is because both referents are salient due to being recently introduced, which gives rise to a violation of informational uniqueness. Recall that personal pronouns require maximal salience, so no problem of this kind arises for them since the more recently mentioned discourse referent is the most salient in this kind of example. (In §5, I will argue that the Northern Khanty Salient Article differs crucially from the Salient Article of Bulu in that the former marks the maximally salient referent.)

Finally, Barlew argues that the Salient Article of Bulu is distinct from the anaphoric articles described by Schwarz and Arkoh and Matthewson [Barlew 2014: 634–636], although he does not provide an in-depth comparison. Part of his argument consists in the observation that neither assumption about what kind of familiarity (weak or strong) an anaphoric article requires would yield the correct distribution of the Salient Article in his examples. For example, if the Salient Article merely required weak familiarity, it should presumably be available in both (21a) and (21b). On the other hand, if it required strong familiarity, it should be barred in both cases. Thus, only a salience requirement captures the observed distribution.

This concludes our discussion of the relevant definite article types.

3.2. Information structure and topicality

Another dimension that is relevant to our discussion of the Salient Article is information structure. After [Lambrecht 1994] and [Nikolaeva 2001], I understand information structure as creating a pragmatically-structured proposition that mediates between the meaning of an utterance and its form.

One part of the proposition corresponds to information assumed by the speaker to be known to the addressee (Lambrecht’s pragmatic presupposition, cf. common ground [Stalnaker 1974]). I will refer to the corresponding part of the utterance as theme, following one terminological tradition [Halliday 1985], in order to avoid confusion with the notion of topic that is used in another sense below.²¹

The other part of the pragmatically structured proposition is defined via Lambrecht’s notion of pragmatic assertion. Pragmatic assertion is that information which the addressee is expected to know upon hearing the utterance. The rheme (focus in Lambrecht’s terminology) is that part of the utterance where the pragmatic assertion differs from the pragmatic presupposition.

To illustrate, consider the example (22) below from the Obdorsk dialect of Northern Khanty. Following standard practice in research on information structure, the theme-rheme division of a given sentence can be inferred from question-answer pairs. The theme of the answer is taken to be the part that is congruent with the question and the rheme is the part that corresponds to the question word (in content questions). Thus, in the answer in (22), the theme is the open proposition “X killed the/a reindeer”, and the rheme is “X = John”, with the latter supplying a value for the variable X in the theme (cf. [Nikolaeva 2001: 4ff.]).

(22) Obdorsk dialect of Northern Khanty

kalaŋ xoj-na we:l-s-a? Juwan-na we:l-s-a

reindeer who-loc kill-pst-pass[3sg] J.-loc kill-pst-pass[3sg]²²

‘Who killed the/a reindeer? John did.’ (adapted from [Nikolaeva 2001: 25])

The reason that I illustrate these notions with data from the Obdorsk dialect is that these data provide a standard of comparison for our discussion of the Kazym dialect Salient Article in the following sections.

In her seminal article [Nikolaeva 2001], Irina Nikolaeva has argued that information-structural notions of primary topic and secondary topic are grammaticalized in the Obdorsk dialect as the syntactic roles of Subject and Agreeing Object (the latter is marked as O2 in the glosses).

Following a long tradition, Nikolaeva defines (primary²³) topic as whatever the proposition is about ([Strawson 1964; Reinhart 1981; Lambrecht 1994] and other works). The aboutness relation holds between a discourse referent and a proposition if the speaker assumes that the discourse referent is the center of current interest about which the assertion is made [Nikolaeva 2001: 4]. By virtue of this definition, the topic is necessarily a part of the theme and is subject to a presupposition of salience (cf. the discussion in §3.1.3 above).

I assume following Roberts [2012; 2019] that each utterance in a discourse answers a (possibly implicit) question under discussion (QUD) and that its theme is defined as that part of the utterance which is congruent with the QUD that it addresses. Given this, the discourse referents figuring in the theme are guaranteed to be relevant and, therefore, salient. Thus, we might add to the definition of the aboutness relation that the referent that the proposition is about figures in the QUD addressed by this proposition, which (at least) partly guarantees its salience.

As for the Obdorsk dialect data, Nikolaeva observes that the constituent that refers to the (primary) topic is required to be the Subject. This leads to extensive use of Passivization to promote topical non-agents to the Subject role. And, conversely, rhematic agents are demoted via the Passive to the Oblique role marked with the Locative -na, given the presence of another topical referent.²⁴ Thus, in (22) the Passive in the answer is obligatory because, based on the question, the patient argument is the topic in this sentence, and the agent argument is in the narrow argument rheme.²⁵

Similarly, in a predicate rheme sentence with the recipient argument as the topic, the recipient is obligatorily promoted to Subject via Passivization, while the agent and patient arguments (in the predicate rheme) are demoted to Obliques.

(23) [“What about Peter?”]

(luw) Juwan-na ke:si-na ma-s-a

he J.-loc knife-loc give-pst-pass[3sg]

‘John gave him a knife.’ (adapted from [Nikolaeva 2001: 25])

Now, Nikolaeva associates another syntactic role with the secondary topic. The secondary topic is defined as “[a]n entity such that the utterance is construed to be about the relationship between it and the primary topic” [Ibid.: 26]. The two referents are presupposed to stand in a relationship. In the Obdorsk dialect, the secondary topic necessarily triggers Object Agreement on the verb, which is the central property of the Agreeing Object (O2) recognized by Nikolaeva as a distinct syntactic role.

As the secondary topic, the O2 is barred from occurring in the rheme, not given by the current QUD. Thus, O2 is excluded from sentences with narrow patient rheme (24), with predicate rheme including the patient (25), or with sentence rheme (26). (The rheme is highlighted in each of these examples.) A Non-agreeing Object (O1) is used in such sentences instead.

(24)

1. [“Whom did John hit?”]
   luw/Juwan Pe:tra re:sk-əs / *re:sk-əs-li
   he/J. P. hit-pst[3sg.o1] / *hit-pst-3sg.o2²⁶
   ‘He/John hit Peter.’
2. Juwan man-e:m wa:nt-əs / *wa:nt-əs-li anta naŋ-e:n
   J. I-acc see-pst[3sg.o1] / see-pst-3sg.o2 not you-acc
   ‘John saw me, not you.’ (adapted from [Ibid.: 29])

(25) [“What did John do?”]

luw Pe:tra re:sk-əs / *re:sk-əs-li

he Peter hit-pst[3sg.o1] / *hit-pst-3sg.o2

‘He hit Peter.’ (adapted from [Ibid.: 30])

(26) [“What happened?”]

ma tam kalaŋ we:l-s-əm / *we:l-s-ə:m

I this reindeer kill-pst-1sg.o1 / kill-pst-1sg.o2

‘I killed this reindeer.’ (adapted from [Ibid.: 28])

Where O2 does occur, it is given in the QUD, so the conditions on secondary topic outlined above obtain. Compare (27) to (25) above. According to Nikolaeva, both examples contain predicate rheme structures, but only in (27) is the Object a secondary topic which ensures obligatory Object Agreement.

(27) [“What did John do to Peter?”]

luw Pe:tra / luw-e:l re:sk-əs-li / *re:sk-əs

he P. / he-acc hit-pst-3sg.o2 / hit-pst[3sg.o1]

‘He hit Peter/him.’ (adapted from [Ibid.])

A similar argument for the role of information structure in the Kazym dialect grammar has been put forward by Natalia Koshkareva [2002] (with important dialectal differences from the Obdorsk dialect). However, in a recent paper, Nikita Muravyev [2022b] has shown that at least in the contemporary Kazym dialect grammar, information structure does not wholly account for the distribution of the Passive and the Object Agreement (see also [Muravyev, Zhornik 2023]).

Rather, topicality hierarchies (person, animacy, definiteness, cf. [Givón 2001a: 156; 2001b: 200; Zúñiga 2006: 21]) play an important role in determining which noun phrase is the Subject or is promoted to Subject and whether the Object is agreed with. This likens the Kazym dialect (and specifically its Passive) to languages with direct/inverse systems [Zúñiga 2006]. Thus, according to Muravyev, only the Passive voice is available in (28) regardless of information structure because the patient mašaj-en ‘Masha’ outranks the agent wot-ən ‘wind’ on the animacy hierarchy (animate > inanimate²⁷).

(28) Kazym dialect of Northern Khanty

1. wot-ən mašaj-en iλ păwət-s-a
   wind-loc М.-poss.2sg down fell-pst-pass[3sg]
   
2. *wot-en mašaj-en iλ păwt-əs
   wind-poss.2sg М.-poss.2sg down fell-pst[3sg]
   ‘The wind knocked Masha off her feet.’ (adapted with translation from [Muravyev 2022b: 51])

Similarly, the person hierarchy constrains the availability of Passive with 1st and 2nd person pronoun agents (speech act participant > third person). As it turns out, these pronouns do not form the Locative Case at all, which is the case assigned to demoted participants, so that they may never be demoted regardless of any other considerations (29). This might be seen as an instance of grammaticalization of a person hierarchy effect: the most prominent noun phrases on the hierarchy cannot be demoted.

(29)

1. ma waśaj-en χătśə-s-əm
   I V.-poss.2sg hit-pst-1sg
2. *waśaj-en <mănɛmən>²⁸ χătśə-s-i
   V.-poss.2sg I.loc hit-pst-pass[3sg]
   ‘I hit Vasya.’ (adapted with translation from [Ibid.: 50])

Another trait common to languages with a direct/inverse system is obviation. Based on the similarities of the Northern Khanty Passive to the Inverse of these languages, Muravyev proposes in an unpublished report [2022a] that the Kazym dialect of Northern Khanty also has an obviation system which, crucially for us, includes the Salient Article. I elaborate on this in the next section.

3.3. Obviation and proximate status

One more parallel that I will consider for the Salient Article is with proximate markers in languages with obviation systems such as the Algonquian languages.

Obviation refers to a separate tier of grammar that distinguishes a privileged third-person argument, called the proximate, from the other third persons, called obviatives. Crucial features of obviation systems [Aissen 2001] involve the condition of “proximate uniqueness” and the condition that only third persons are eligible for proximate/obviative marking, with the first- and second-person pronouns excluded from consideration. Proximate uniqueness states that there must be one and only one proximate in an obviation span (which might consist in a single clause or a whole episode of a narrative) [Aissen 2001: 12].

Obviation systems are a common trait of Algonquian languages but are also found in the other language families of the Americas, as well as beyond that region (see [Ibid.] for a preliminary typology). In Algonquian languages, assignment of proximate status is conditioned on the following hierarchies [Ibid.: 6–7]:

• animacy (Anim > Inan, i.e., animates are preferred as proximates);
• semantic role (Agent > Patient);
• topichood (Top > Non-top);
• possessorhood (Possessor > Possessee, in a possessive construction the Possessor is preferred as proximate over the Possessee).

Beyond that language family, mostly the same constraints are relevant, although particular languages may diverge from this set to varying degrees, and some systems also include pronominality (Pron > Common N) as a relevant hierarchy.

As mentioned above, Muravyev [2022a] hypothesizes that the Kazym dialect also has an obviation system which includes the following grammatical devices distinguishing the proximate from obviatives:

1. The Passive marks sentences with proximate patients (28);
2. The Salient Article marks proximate status;
3. Possessive agreement is triggered by proximate lexical Possessors.²⁹

For the purposes of this paper, I forgo a deeper discussion of the Passive and the Possessive agreement, focusing on the Salient Article. And in the remainder of this section, I introduce Heather Bliss’ [2017] analysis of obviation in Blackfoot (< Algonquian) to serve as a standard of comparison when we consider Muravyev’s hypothesis regarding the Salient Article in §§4.3–4.4.

In Blackfoot, both Proximate and Obviative receive special morphological marking (30). As expected, proximate uniqueness constrains occurrence of more than one proximate per obviation span (here, clause) (31).

(30) Ann-wa Leo ííhpok-inihkim-yii-wa ann-yi n-Itán-yi

dem-prox L. accom-sing.ta-dir-prox dem-obv 1-daughter-obv

‘Leo sang with my daughter.’ (adapted from [Bliss 2017: 4])

(31) *Ann-wa Leo ííhpok-inihkim-yii-wa ann-wa n-Itán-wa

dem-prox L. accom-sing.ta-dir-prox dem-prox 1-daughter-prox

Intended: ‘Leo sang with my daughter.’ (adapted from [Ibid.])

Bliss’ paper is concerned with the question of whether obviation in Blackfoot is a primarily syntactic or discourse phenomenon, which has been the subject of intense debate in Algonquian studies (see the summary and references in [Ibid.]). Bliss argues that Blackfoot Proximate and Obviative categories code primarily syntactic distinctions, but the nature of these distinctions makes these categories suitable for “recruitment” for discourse purposes such as main character vs. peripheral character marking in narratives [Goddard 1990].

According to Bliss, the core syntactic property coded by Blackfoot Proximate/Obviative marking is syntactic (in)dependence. Proximate nominals exhibit free word order (32) and may be omitted (33). (Examples below are adapted from [Ibid.: 7–8].)

(32)

1. Ann-wa n-ínsst-innaan-wa nít-sspommo-a-wa
   dem-prox 1-sister-1pl-prox 1-help.ta-dir-prox
   ‘I helped our sister.’
2. Nítsspommoawa anna nínsstsinaana³⁰
   Ibid.

(33)

1. A’páwaawahkaa-wa ann-wa Pióhkomiaaki
   walk.around.ai-prox dem-prox far.sounding.woman
   ‘Far Sounding Woman is walking around.’
   
2. A’páwaawahkaawa
   ‘S/he is walking around.’

On the other hand, the linear position of Obviative nominals is constrained: they cannot occur preverbally (unless resumed by an enclitic pronoun):

(34)

1. Áókataki-yini ann-yi w-inssts-yi
   bead.ai-obv dem-obv 3-sister-obv
   ‘His sister does beadwork.’
2. *Anni ónssts áókatakiyini
   Ibid.

Furthermore, Obviatives cannot be freely omitted (similarly, unless resumed by an enclitic pronoun):

(35)

1.  Áísokssta-yini ann-yi ot-issítsimaan-yi
   nurse.well.ai-obv dem-obv 3-baby-obv
   ‘Her baby is nursing well.’
2. Áísokssta-yini-áyi
   nurse.well.ai-obv-3sg.prn
   ‘S/he is nursing well.’
3. *Áísoksstayini.
   Ibid.

Bliss suggests that the reason for these distributional differences between Proximates and Obviatives in Blackfoot is that Proximate nominals are adjuncts to the clause, while Obviative nominals are arguments. This directly reflects the intuition that Proximates are syntactically independent, and vice versa for Obviatives.

This claim is supported by the fact that Proximate-marked verbs and nouns can function as independent clauses, while Obviative nominals necessarily function as arguments (dependent on the verb), and the same is true of Obviative-marked verbs which must appear as dependent clauses. See the examples in [Bliss 2017: 10–12].

Having argued for the syntactic nature of obviation in Blackfoot, Bliss also discusses its discourse function. She considers a narrative and shows that Proximate marks the main character in this narrative [Ibid.: 12−15].³¹ Later in the story, when the main character of the story changes, a different noun phrase receives Proximate marking, while peripheral characters are marked Obviative (cf. [Goddard 1990]). This clearly shows that obviation devices can be intentionally deployed by narrators to highlight referents that are central to their story. As mentioned above, Bliss views this as an instance of a recruitment process whereby a syntactic device is utilized for a discourse function.

Furthermore, Bliss suggests that the “Proximate as independent/Obviative as dependent” analysis extends to other obviation systems in Algonquian, citing evidence from a range of other Algonquian languages. For instance, Bliss cites Benjamin Bruening’s analysis of obviation in Passamaquoddy [Bruening 2001; 2009]. Bruening proposes that the Proximate/Obviative distinction works similarly to the Nominative/Accusative distinction in dependent case theory [Marantz 1992] where Accusative is a dependent case assigned to a noun phrase in the presence of another structurally prioritized NP (see §4.4 for details). The Obviative is similarly dependent in Passamaquoddy, as it is assigned in the presence of a Proximate NP. Thus, according to Bliss, obviation in Algonquian is a syntactic phenomenon that rests on syntactic (in)dependence, realized differently in different languages.

When we discuss Northern Khanty data, I will argue against a syntactic account of the Salient Article’s distribution (§4.4) despite its similarities to proximate markers in Algonquian.

4. The Salient Article as a definite determiner

Having introduced the relevant theoretical notions and cross-linguistic “standards of comparison”, we are now prepared to consider the Salient Article in detail. In Section 4.1, I show that the Salient Article requires both uniqueness and familiarity — the defining property of definite determiners (§3.1) — but these conditions are not sufficient. In Section 4.2, I argue that the availability of the Salient Article does not depend on information-structural distinctions or on syntactic role of the NP. In Sections 4.3 and 4.4, I consider the Salient Article as a proximate marker and argue against a syntactic analysis.

4.1. Uniqueness and familiarity

The Salient Article demonstrates some of the basic properties of definite determiners. Consider the marking possibilities of noun phrases referring to the soldier in a small narrative-like sequence in the following three examples (based on the example from [Nikolaeva 2003: 139–140]).

(36) ma χot-a λuŋ-s-əm. śăta săldat-(^#en) oməs-λ.

I house-dat enter-pst-1sg there.loc soldier-poss.2sg sit-npst[3sg]

‘I entered a house. There was a soldier sitting there.’

(37) ma săldat-(^%en) χuś-a wana măn-s-əm, puškan-ən šăš-s-ɛm

I soldier-poss.2sg at-dat closer go-pst-1sg gun-loc show-pst-1sg>sg

‘I came closer to the soldier and raised my gun at him.’

(38) săldat-^#(en) pakn-əs.

soldier-poss.2sg get.frightened-pst[3sg]

‘The soldier got frightened.’

Upon first mention of the soldier (36), the Salient Article is barred. Upon second mention (37), where the Subject is still the narrator, and the soldier noun phrase is in an Oblique position (complement of a postposition), the bare form is preferred, although the Salient Article is at least acceptable for some speakers (signified by the percent sign). Finally, in (38), where the soldier is the Subject, the Salient Article is obligatory for all the speakers I consulted.

These data show that the Salient Article is unacceptable with novel referents and that it can be used with familiar referents. Based on this, we may propose that the Salient Article requires familiarity, although it does not seem to be a sufficient condition, as suggested by (37).

In the case of novel unique referents, however, the Salient Article is unacceptable. For instance, it is generally not accepted with superlatives (39), unlike the English Definite Article (17). This suggests that uniqueness is not a sufficient condition for the Salient Article to be used.

(39) [There is a fishing contest on Fisherman’s Day. I come home from the contest and tell my family about it.]

tăm oλ mɛt wɵn χʉλ-(^#en) wet-χuś-jăŋ kila-kɛm

this year very big fish-poss.2sg five-dec-ten kg-appr

‘This year, the biggest fish was 15 kilograms.’

Let us demonstrate that the Salient Article indeed requires informational uniqueness / maximality. Example (40a) shows that the Salient Article is acceptable with maximal plural referents that were introduced in prior context. In the same context, a non-maximal NP does not admit the Salient Article (40b). The Partitive Article poss.3sg must be used instead.³²

(40) [“I was walking along the street when I saw two dogs.”]

1. amp-ŋəλ-an ma pɛλ-am-a χurət-ti pit-s-əŋən
   dog-du-poss.2sg I towards-poss.1sg-dat bark-nfin.npst become-pst-3du
   ‘The dogs started barking at me.’
2. i amp-əλ/^#-en ma pɛλ-am-a
   one dog-poss.3sg/-poss.2sg I towards-poss.1sg-dat
   χurət-ti pit-əs
   bark-nfin.npst become-pst[3sg]
   ‘One of the dogs started barking at me.’

This suggests that the Salient Article presupposes uniqueness which gives rise to infelicity in (40b).

Further support for this proposal comes from the obligatory use of the Salient Article in examples like (41). Here, the bare form is barred because, judging by my consultants’ comments, it gives rise to an anti-familiarity/anti-uniqueness inference, i.e., the dog referred to is not the unique familiar dog introduced in prior context. (Cf. the obligatoriness of the English Definite Article in (17) from §3.1.)

(41) [“I was walking along the street when I saw a dog.”]

amp-^#(en) ma pɛλ-am-a χurət-ti pit-əs

dog-poss.2sg I at-poss.1sg-dat bark-nfin.npst become-pst[3sg]

‘The dog started barking at me.’

Consultant’s comment on Ø: “then it’s some other dog, not clear which”.

One way to account for such inferences is by appealing to Heim’s pragmatic principle Maximize Presupposition! [Heim 1991], which enforces the use of definite articles in contexts where their (uniqueness and familiarity) presuppositions are satisfied. According to my proposal (presented in §5), the Salient Article presupposes informational uniqueness and weak familiarity (in Roberts’ sense (14)), and failing to use it in a context where its presuppositions are satisfied leads the addressee to infer that its use would not be felicitous in this context, e.g., because the intended referent is not unique/familiar.

So far it seems that the Salient Article is much like the English Definite Article or, perhaps, the narrower Akan Definite Article. Is our job here done No, because the Salient Article is still narrower than both these markers! It differs from uniqueness-based definites in that it is not used with superlatives (39) and it differs from anaphoricity-based definites in that strong familiarity is not a sufficient condition on its use (e.g., in (37) the bare form is preferred with a strongly familiar referent).

Example (42), repeated with adjustments from (7), also shows that strong familiarity is not enough. Here a passport was introduced in the preceding context, which makes it strongly familiar (even if the story digresses between the two mentions of the passport). Nevertheless, the Salient Article is excluded here, and the Associative poss.1sg must be used instead (cf. §2.3). Thus, strong familiarity and uniqueness cannot be all there is to Salient Article’s conditions of use.

(42) [“I found somebody’s passport in the street. I went to the town administration. Met a friend there and talked to her for some time.”]

nɛm nɛpek-ɛm/^#-en suwet-ən χăj-s-ɛm

name paper-poss.1sg/-poss.2sg council-loc leave-pst-1sg>sg

‘[Then] I left the passport at the administration. [Let them find the owner.]’

One candidate for another condition on the Salient Article is the requirement of salience, as proposed by Barlew [2014] for the Salient Article of Bulu (§3.1.3). For example, it seems reasonable to assume that one important difference between (42), where the Kazym Khanty Salient Article is barred, and (41), where it is obligatory, is that in the latter the referent is salient to a high degree but in the former it is not.

However, a requirement of salience (of some positive degree) by itself, which was enough for the Bulu Salient Article, still wouldn’t be sufficient for the Kazym Khanty Salient Article. Recall that according to Roberts and Barlew a factor determining salience is relevance of the referent to the current question under discussion (QUD). Even though the passport cannot be assumed to be at the center of the addressee’s attention in the context immediately preceding (42), the QUD “What did you do with the passport ” presumably still looms somewhere in the stack of relevant questions under discussion at that point, which ensures that the passport is salient to some positive degree, even if not to a high one. The same applies to the narrative examples (36)–(38): the soldier referent arguably becomes relevant to the story upon his introduction in (36) which guarantees that he is salient to some degree. Nevertheless, the Kazym Khanty Salient Article is still dispreferred upon second mention of the soldier in (37) and is barred in (42), which one may attribute to the fact that the soldier has not yet attained a sufficiently high degree of salience and the passport has already lost it by the time it is mentioned for the second time.

What this discussion suggests is that, perhaps, unlike the Salient Article of Bulu [Barlew 2014: 630–631], the Kazym Khanty Salient Article does care to which degree the NP referent is salient. Before we develop this idea in some detail, let us consider some further possible factors.

One candidate suggested in [Mikhailov 2021a] is information structure. There, I proposed that the Kazym Khanty marker is a “topic marker”. Following this suggestion, it is reasonable to check if perhaps it marks topics in Nikolaeva’s [2003] sense Another candidate is syntactic role. In all the examples above, except the one where the Salient Article is unavailable (42), the NP with the Salient Article is the Subject. I address these two possibilities in the following section.

4.2. Information structure and syntactic role

In the preceding section I argued that the Salient Article is a definite determiner which must have some additional requirements besides informational uniqueness and existence. In this section I show that neither syntactic role nor information structure provide the needed requirements of the Salient Article.

Let us consider a set of examples from an elicited dialogue about a thief stealing a computer from the local museum. Examples in (43) show that the Salient Article is used with both Subjects and Objects. (Note that the context in these examples ensures that the thief is topical and in the theme of the utterance, as defined in §3.2.)

(43) [“— A computer was stolen from our museum. — Did they find the thief yet ”]

1. pa muj, muλχatəλ λoλmaχ-en wɵjət-s-a,
   add what yesterday thief-poss.2sg find-pst-pass[3sg]
   kampjuter-ew juχλi mă-s-i
   computer-poss.1pl back give-pst-pass[3sg]
   ‘You betcha! Yesterday, the thief was found, and the computer has been returned.’
2. pa muj, ma muλχatəλ λoλmaχ-en wɵjət-s-ɛm
   add what I yesterday thief-poss.2sg find-pst-1sg>sg
   ‘You betcha! Yesterday, I found the thief.’

The Salient Article may also be used with non-core syntactic roles such as Demoted Agent in the Locative (44) and Possessor (45).

(44) [“— A computer was stolen from our museum. — Did they return the computer yet ”]

kampjuter-ew λoλmaχ-^#(en)-ən tinij-əm

computer-poss.1pl thief-poss.2sg-loc sell-nfin.pst[pass.3sg]

‘It turns out that the computer was sold somewhere by the thief.’³³,³⁴

(45) [“— A computer was stolen from our museum. — Did they find the thief yet — You betcha! Yesterday, I found the thief. — What did they do to him in the end ”]

λoλmaχ-en mašinaj-əλ milcaj-ən nuχ wʉ-s-i

thief-poss.2sg car-poss.3sg police-loc up take-pst-pass[3sg]

‘The thief’s car was taken by the police. [And he himself was put to jail.]’

I conclude from these data that the Salient Article has no special restrictions on syntactic role.

Now we turn to (the lack of) information-structural restrictions. Recall from §3.2 that in the Obdorsk dialect of Northern Khanty there is a strict link between primary topics and the Subject position and secondary topics and the Agreeing Object (O2) position as argued by [Nikolaeva 2001]. Thus, in agent rheme sentences primary topic patients are obligatorily promoted to Subject via Passivization (and rhematic agents are obligatorily demoted). Narrow argument rheme on the Subject or the O2 is impossible. In other words, in utterances that do have an explicit theme, the Subject NP is obligatorily a part of the theme (cf. footnote 24). If another NP is also topical, it is obligatorily in the O2 position.

Is there such a strict correspondence between information structure and the Salient Article in the Kazym dialect I argue no.

In most of the examples considered so far, the NP with the Salient Article was indeed either a primary or a secondary topic. However, example (44) already shows that the Salient Article may be obligatory even with a rhematic NP.

To this we may add example (46) where a NP is under narrow argument rheme and is nevertheless obligatorily marked with the Salient Article.

(46) [We are putting on a play at the school theatre. Two girls have not yet been assigned a role. We discuss their strengths to decide who will play the sun: “— The girl on the left is tall. — The girl on the right is a good singer. So, who will play the sun ”]

jăm pɛlək-ən oməs-ti ew-^#(en) at junt-λ

good side-loc sit-nfin.npst girl-poss.2sg opt play-npst[3sg]

‘Let the girl on the right play [the sun].’

The same observation may be drawn from example (47) where a demonstrative NP is under narrow contrastive rheme and is also obligatorily marked with the Salient Article.

(47) [I point at a boy and ask: “Who’s that ” My addressee answers: “This boy’s name is Petya.” Our friend corrects them.]

śi aj_ik-^#(en) χɵn petˊa, tum aj_ik-en petˊa

dem boy-poss.2sg neg.emph P. that boy-poss.2sg P.

‘This boy is not Petya, that one is.’

Table 2 summarizes the distribution of the Salient Article with respect to information structure as compared to Obdorsk dialect Subjects in argument or predicate rheme utterances (cf. Section 3.2).

Table 2. Independence of the Kazym dialect Salient Article from information-structural status as compared to the Obdorsk dialect Subjects and Agreeing Objects (O2) [Nikolaeva 2001]

Thus, I suggest that whether an NP is in the theme or in the rheme of the utterance is irrelevant for the Salient Article as long as its referent itself is salient in the given context. In all the examples in this section where the Salient Article is obligatory, the referent of the marked NP is salient by virtue of the preceding context: the referent is either the primary topic of the target utterance, or it has been the primary topic of the preceding utterances (ensuring salience), or its salience is derived from bridging to another topic of the stretch of discourse in question.

The only example where I marked the Salient Article as optional — the second sentence of the soldier example (37) — was an example where the context is underdetermined with respect to salience of the target referent and, thus, there is some ambiguity as to the speaker’s intentions: whether the speaker intends the addressee to attend to the referent in question to a significant degree or not. I consider this to be the reason for the Salient Article’s optionality in cases like this one.

4.3. Obviation

Finally, let us consider Muravyev’s [2022a] hypothesis that the Salient Article is a proximate marker much like those found in Algonquian languages (see §3.3).

Recall that two of the core properties of obviation systems is the proximate uniqueness constraint [Aissen 2001] and the condition that speech act participants are not considered for proximate/obviative marking. Both properties turn out to be observed by the Salient Article.

Firstly, only one third person NP may be marked with the Salient Article within a single clause, as (48) and (49) show. In both examples, the less topical of the two NPs must be marked with the Associative poss.3sg instead (cf. §2.3). In (48), the boy is associated with the girl via the preceding context. In (49), one may argue that the thief is associated with the computer under discussion since he stole it. What is crucial for our purposes is that the Salient Article is unavailable for both these referents, even though the boy is the secondary topic in (48) and the thief is salient in (49) (cf. footnote 34).³⁵

(48) [“I was walking along the street when I saw a boy and a girl.”]

ew-en aj_ik-eλ / ^#aj_ik-en χătś-əs-λe

girl-poss.2sg boy-poss.3sg / boy-poss.2sg hit-pst-3sg>sg

‘The girl hit the boy.’

(49) [“A computer was stolen from our museum. But everything worked out.”]

waśaj-en λoλmaχ-əλ/^#-en wɵjət-s-əλe

V.-poss.2sg thief-poss.3sg/-poss.2sg find-pst-3sg>sg

‘Vasya found the thief [and returned the computer].’

Compare (49) with the minimally different (50) where the other NP is an indefinite pronoun. An indefinite referent is obviously less salient than the thief, so the thief is marked with the Salient Article as the proximate third person.³⁶

(50) [“A computer was stolen from our museum. But everything worked out.”]

λoλmaχ-en χuj-at-ən wɵjət-s-a,

thief-poss.2sg who-indef-loc find-pst-pass[3sg]

‘Somebody found the thief [and returned the computer].’

Secondly, if the first NP refers to an interlocutor in the thief examples, the second NP must be marked with the Salient Article (51) (repeated from (43)). Just as expected if the Salient Article marks the proximate third person.

(51) [“— A computer was stolen from our museum. — Did they find the thief yet ”]

pa muj, ma muλχatəλ λoλmaχ-en wɵjət-s-ɛm

add what I yesterday thief-poss.2sg find-pst-1sg>sg

‘You betcha! Yesterday, I found the thief.’

So far, the Salient Article fits the profile of proximate marking perfectly. The natural question arises: should the Salient Article be analyzed in a similar way to proximate markers in Algonquian languages (cf. §3.3), i.e., as a marker of syntactic independence [Bliss 2017]? In the following section I argue against such an analysis.

4.4. Against a syntactic account of the Salient Article

Recall from §3.3 that Heather Bliss [2017] proposed to analyze proximate markers in Algonquian languages as markers of syntactic independence (and obviative markers as markers of syntactic dependence). According to Bliss, this distinction is realized differently in different obviation systems across the language family, e.g., in Blackfoot syntactic independence corresponds to adjunct/independent constituent status, while in Passamaquoddy it roughly corresponds to the nominative case as opposed to the accusative [Bruening 2001].

The question is whether we should account for the distribution of the Salient Article observed in the preceding section in a similar way. I would like to argue against this line of analysis.

First, consider again the generalizations about the Salient Article that we discussed so far. In §4.2, I have argued that the Salient Article is not restricted to any particular syntactic role(s). In general, I have found no special syntactic restrictions on NPs with the Salient Article as compared to NPs without it found within the same clause. If the presence/absence of the Salient Article were to correspond to the Algonquian proximate/obviative distinction (under Bliss’ analysis), we should find an underlying syntactic distinction that would drive the salience effects observed in the preceding sections. In my current data, there is no evidence to that effect.

Second, there are data that would significantly complicate a syntactic account. As we have seen above, the Salient Article is also used with NPs with demonstratives. It turns out that with demonstrative NPs, the Salient Article is no longer subject to proximate uniqueness. Thus, in (52), two demonstrative NPs with the Salient Article are perfectly felicitous within the same clause. Compare (48) above where two NPs with the Salient Article are infelicitous without demonstratives. (This observation is due to Nikita Muravyev [2022a].³⁷)

(52) tăm ik-en tum im-en šiwəλ-əs

this man-poss.2sg that woman-poss.2sg notice-pst[3sg]

‘This man saw that woman.’ (adapted from [Muravyev 2022a: ex. 9])

To see why this observation presents a problem for syntactic accounts, let us consider two possible lines of analyses in more detail.

The literature presents at least two syntactic approaches to proximate uniqueness. The first one is the optimality-theoretic (OT) account by Judith Aissen [2001].³⁸

According to Aissen, obviation can be modeled via a set of markedness constraints which ban the occurrence of either obviative third person NPs (*Obviative), proximate third person NPs (*Proximate) or two (or more) proximates (Proximate Uniqueness) within a single obviation span [Ibid.: 10–12]. The generator component of Aissen’s model generates a candidate set that contains tuples of third persons NPs each assigned either proximate or obviative status. Next, the evaluator component applies to the candidates the hierarchy of constraints to find the optimal candidate. The hierarchy of constraints that correctly rules out two proximates within a single obviation span is given in (53).

(53) An optimality-theoretic hierarchy of constraints for obviation (based on [Aissen 2001])

Proximate Uniqueness ≫ *Obviative ≫ *Proximate

To see how this works, consider Tableau 1 that models a basic example of proximate uniqueness effects, e.g., our example (48) above.

Tableau 1. Proximate Uniqueness in Optimality Theory (based on [Aissen 2001: 12])

The generator component supplies three candidates: (a) with two proximates, (b) with two obviatives, and (c) with one proximate and one obviative. In Northern Khanty terms, this would, roughly speaking, correspond to (a) a clause with two NPs with the Salient Article, (b) a clause with no NPs with the Salient Article, (c) a clause with one NP with the Salient Article.³⁹ The (a) candidate is ruled out since it violates Proximate Uniqueness, the highest ranked constraint. And the optimal candidate is (c), which, unlike (b), violates the second highest ranked constraint only once. Thus, Aissen’s account correctly derives the observed pattern in cases without demonstratives.

It is not clear, however, how this analysis could be extended to cases with demonstratives like (52). One possibility is to say that NPs with demonstratives form an independent obviation span. This proposal seems problematic. There does not seem to be any reason why demonstrative NPs would form a separate obviation span. After all, obviation spans are clauses or larger units of discourse (as per Aissen’s definition [1997: 713–714] which is based on much preceding work on Algonquian). There do not seem to be cases where an obviation span is smaller than a clause. Aissen even states explicitly that “it seems impossible to initiate a new obviation span within a clause” [Ibid.].

Therefore, it seems that an Aissen-style OT account of obviation effects in the Salient Article would require a modification of the notion of obviation span that is both ad hoc and contradicts the Algonquian data, which is why I will not pursue this line of analysis any further.

The second syntactic approach to proximate uniqueness effects was proposed by Bruening [2001] for Passamaquoddy within a Minimalist Program (MP) generative syntactic framework and later applied to Tundra Nenets (< Uralic) data by [Bárány, Nikolaeva 2021] (cf. footnote 29).

In Bruening’s system [2001: 119ff.], all NPs bear a syntactic proximate feature [P] which may be either valued [+P] or unvalued [P]. First and second persons are inherently [+P]. Third persons are not inherently valued but may become valued [+P] from context and in opposition to another third person NP. If two third person NPs co-occur, one will become [+P], while the other will remain unvalued [P] and receive an additional obviative feature [Obv] which is spelled out as obviative morphology. (Note that in Passamaquoddy there is no proximate morphology, unlike Blackfoot.)

Both variants of the proximate feature [+P]/[P] are uninterpretable and need to be checked off by functional heads in the syntax (which is a standard set of assumptions in MP generative syntax): [P] is checked off by the v0 (voice) head, while [+P] must be checked off by some higher head which Bruening calls H0. This essentially means that all NPs need to be licensed in a particular structural configuration. In Passamaquoddy, this setup allows to model movement of the proximate to the left of the verb and agreement of the verb with the proximate.

More formal detail than that need not concern us here, but the general idea is that proximate status of a third person NP may only be received in competition with another third person NP. This entails that the only third person that co-occurs with a second or first person will be neither proximate nor obviative in Bruening’s model. (Recall that the proximate does not receive any overt morphological marking in Passamaquoddy.) And there is a good reason for this assumption with respect to Passamaquoddy data.

In a Passamaquoddy sentence with a first or second person and a third person, only the former is indexed by an agreement prefix on the verb and appears to the left of it. In a sentence with two third persons, either one may be indexed with the same agreement prefix and appear to the left of the verb. Thus, it seems that the bare third person NP in the first case indeed is not proximate as it cannot be indexed by the agreement prefix and appear to the left of the verb. On the other hand, it is also not obviative, as it does not bear obviative morphology. This pattern may be summarized in terms of features as follows:

(54) Passamaquoddy obviation in Bruening’s model [Bruening 2001: 119–124]

(a) 1st/2nd person proximate & 3rd person non-proximate: [1/2, +P] … [3, P]

(b) 3rd person proximate & 3rd person obviative: [3, +P] … [3, P, Obv]

At this point, it is already obvious that Bruening’s approach does not derive correct predictions for the Salient Article, since the Salient Article freely appears in the (a) case, see (55) repeated from (51).

(55) [“— A computer was stolen from our museum. — Did they find the thief yet ”]

pa muj, ma muλχatəλ λoλmaχ-en wɵjət-s-ɛm

add what I yesterday thief-poss.2sg find-pst-1sg>sg

‘You betcha! Yesterday, I found the thief.’

If the Salient Article were associated with proximate status, which corresponds to the [+P] feature in Bruening’s model, then it must have been barred in (55) according to (54), which is not the case. If we were to allow [+P] to be freely assigned to any third person, we would lose Bruening’s account of proximate uniqueness which is wholly grounded in competition between two local third persons.

On the other hand, it seems easy to account for violations of proximate uniqueness with demonstrative NPs (52) in Bruening’s model. One has only to assume that demonstratives are inherently valued [+P], just like first and second person pronouns. This, of course, is stipulative but at least does not require any additional assumptions beyond the stipulations already made by Bruening himself.

Still, Bruening’s approach to proximate uniqueness makes wrong predictions in the case of (55), which is why I also reject it as a model of the Salient Article.

Another problem with both the OT and the MP approaches is that they have nothing to say about the Salient Article’s status as a definite determiner, which I argued for in §4.1. In the next section I sketch a semantic analysis of the Salient Article that both reflects its status as a definite and makes correct predictions for its distribution with respect to proximate uniqueness without unmotivated stipulations.

5. Toward an analysis

Let us sum up what generalizations an analysis of the Salient Article must account for:

1. The Salient Article is a definite determiner as it requires (at least) informational uniqueness and familiarity in the sense of [Roberts 2003] but it does not fit the general profiles of uniqueness and familiarity-based determiners (§4.1).
2. The Salient Article is used with demonstratives (examples (2), (6), (47)).
3. The Salient Article is subject to a kind of proximate uniqueness constraint [Aissen 2001] such that there may be at most one NP with the Salient Article per clause (§4.3).
4. If a NP with the Salient Article has a demonstrative, it is no longer subject to proximate uniqueness (§4.4).

In the preceding section, I argued that neither of the two existing syntactic accounts of proximate uniqueness work for the Salient Article. Thus, the Salient Article cannot be reasonably said to be syntactically proximate in any sense. Instead, I propose a semantic account based on the intuition voiced in the preceding sections that the Salient Article requires salience of the discourse referent in question.

Recall from §§3.1–3.2 that we understand salience as a function of attention capture [Roberts 2011; Barlew 2014]. The salient discourse referents are those that the interlocutors attend to, be that in virtue of relevance to the interlocutors’ goals or due to the referents’ perceptual prominence. Assuming that the Salient Article requires salience as well as informational uniqueness and weak familiarity, this already accounts for the data in §§4.1–4.2.

Furthermore, following general consensus in the work on demonstratives, in particular [Diessel 2006],⁴⁰ I assume that demonstratives serve to coordinate the interlocutors’ joint attention. Speakers use demonstratives and deictic pointing gestures to focus the addressee’s attention on a particular referent. Given that the demonstrated referent comes into addressee’s attention, this ensures that the referent is salient. And this invites the use of the Salient Article as the demonstrative fulfils the Salient Article’s salience requirement on the discourse referent.

Thus far we have said nothing to account for the obviation facts. I propose that the Salient Article requires maximal salience of the discourse referent among the salient discourse referents in a given context. This proposal follows in spirit the discourse-based approaches to the Algonquian proximate category which state that the proximate marks the more salient discourse referent among the referents under discussion (see the references in [Aissen 1997; Bliss 2017]). My proposal for the semantics of the Salient Article, based on Roberts’ analysis of definites (14), is summarized in (56) (with the added part highlighted).

(56) Semantics of the Northern Khanty Salient Article (first version)

Given a context C, the use of a NPi with the Salient Article presupposes that it has as antecedent a discourse referent xi which is:

1. weakly familiar in C,
2. unique among discourse referents in C in being contextually entailed to satisfy the descriptive content of NPi, and
3. the most salient discourse referent in C excluding interlocutors.

Let us discuss it in detail.

The difference between my proposal and Roberts’ analysis lies in the condition (c) which accounts for the proximate uniqueness effects. Consider again (48) repeated as (57). Here, only the Subject NP ew-en ‘girl’ may be marked with the Salient Article even though the preceding context does not make either the girl or the boy more salient.⁴¹

(57) [“I was walking along the street when I saw a boy and a girl.”]

ew-en aj_ik-eλ / #aj_ik-en χătś-əs-λe

girl-poss.2sg boy-poss.3sg / boy-poss.2sg hit-pst-3sg>sg

‘The girl hit the boy.’

I suggest that in cases like this the speaker intends one referent (the girl) to be more salient than the other one and uses the Salient Article, which presupposes maximal salience of this referent, to convey this intention to the addressee. Seeing how the Salient Article’s presupposition is not satisfied in the context but is nevertheless plausible, the addressee accommodates it and shifts the focus of her attention to the referent marked with the Salient Article.⁴²

Since only one referent may be maximally salient in a particular context at a particular point in time, marking the second referent (the boy) with the Salient Article would lead to presupposition failure. This gives the observed infelicity of using the Salient Article two times per clause in (57). Thus, my proposal accounts for the proximate uniqueness effects observed in the Salient Article.

Finally, consider again the case with two demonstrative NPs with the Salient Article (repeated from (52)):

(58) tăm ik-en tum im-en šiwəλ-əs

this man-poss.2sg that woman-poss.2sg notice-pst[3sg]

‘This man saw that woman.’ (adapted from [Muravyev 2022a: ex. 9])

What would a typical utterance of a sentence like (58) look like The speaker would first point at the man, uttering tăm iken ‘this man’, and then at the woman, uttering the rest of the sentence tum imen šiwəλəs ‘saw that woman’. The addressee would follow with her gaze the first pointing gesture to the man and then the second pointing gesture to the woman. Thus, it is fair to say that at different moments during the locutionary act of (58) two different referents are maximally salient, i.e., in focus of the interlocutors’ joint attention. And this is why the use of two Salient Articles per clause is possible with demonstratives: demonstratives allow for the interlocutors’ joint attention to shift during the utterance.

I amend my semiformal analysis in (59) to explicitly mention moments of time that maximal salience of the referent is evaluated at.

(59) Semantics of the Northern Khanty Salient Article (final version)

Given a context C and time t, the use of a NPi with the Salient Article presupposes that it has as antecedent a discourse referent xi which is:

1. weakly familiar in C,
2. unique among discourse referents in C in being contextually entailed to satisfy the descriptive content of NPi, and
3. the most salient discourse referent in C at t excluding interlocutors (where t may be either the utterance time tUT or the time of the pointing gesture tDEM when a demonstrative is present).

To conclude the presentation of my analysis, let me discuss the final part of the condition in (c). As per (c), the Salient Article is only concerned with salient discourse referents to the exclusion of the interlocutors. This is a necessary condition since it ensures that the Salient Article is still available when the other NP is first or second person (cf. §§4.3–4.4). But is there any motivation for excluding the interlocutors or is it an ad hoc condition that we cannot do without I would like to suggest that there is motivation for this assumption.

One of the existing views on the semantics of first and second person pronouns⁴³ treats them as indexicals that refer directly to the speaker and the addressee of the given context and not to the available discourse referents. Under this view, the context is understood as a tuple consisting of the interlocutors, the common ground, the set of discourse referents, and the utterance time.⁴⁴ Unlike third person pronouns and noun phrases that are interpreted as variables denoting discourse referents, the first and second person pronouns are interpreted as constants referring directly to the interlocutors. Thus, under the indexical view, the interlocutors are not even contained in the set of weakly familiar discourse referents and will not be considered for the use of the Salient Article as it is already restricted to discourse referents. Strictly speaking, we can do without the addition to condition (c) that excludes the interlocutors, but I leave it for the sake of explicitness.⁴⁵

To conclude this section, I have proposed a semantic analysis of the Northern Khanty Salient Article that correctly accounts for the nuances of its distribution outlined in the preceding sections and does so without unmotivated stipulations.

6. Conclusion

I have argued that the Salient Article is a definite determiner that marks the most salient discourse referent in the given context. As such, it represents a hitherto unattested definite type, different from uniqueness-based and familiarity-based definites [Schwarz 2009; Arkoh, Matthewson 2013] and from the Salient Article of Bulu [Barlew 2014].

In pursuing this analysis, I have considered and rejected syntactic alternatives based on the similarity of the Salient Article to the proximate markers of the Algonquian languages. The argument rested on applying to the Salient Article two existing syntactic analyses of proximate uniqueness effects [Aissen 2001; Bruening 2001]. This yielded wrong predictions, so the analyses were rejected. It would be interesting to see whether the semantic analysis of the Salient Article proposed here may be productively applied to (any of) the Algonquian data, which is, however, a task for future work.

If the analysis proposed here is on the right track, it lends further support to the view that salience, understood as a function of the interlocutors’ attentional states, is a crucial dimension for the semantics of definite expressions, as argued by [Gundel et al. 1993; Barlew 2014; Roberts 2022], among others.

Finally, as I mentioned in the introduction, there has recently been a trend in Uralic linguistics set by [É. Kiss 2018; É. Kiss, Tánczos 2018; Halm 2018; Serdobolskaya et al. 2019] of analyzing non-possessive uses of Uralic possessives as independent markers in a theoretically informed fashion. The current paper supports this trend, with the hope that more work of this kind will follow.

ABBREVIATIONS

1, 2, 3 — first, second, third person

ACC — accusative case

ACCOM — accompaniment

ADD — additive particle

AI — animate intransitive

APPR — approximative

COP — copula

DAT — dative case

DEC — numeral suffix for 11–17

DEM — demonstrative

DIR — direct

DU — dual number

EMPH — emphatic particle

FAM — Akan Definite Article

INDEF — indefinite

LOC — locative case

NFIN — general nonfinite form

NPST — non-past tense

NSG — non-singular number

O1/O2 — Obdorsk Non-agreeing/Agreeing Object

OBV — obviative

OPT — optative mood

PASS — passive voice

PAST, PST — past tense

PL — plural number

PN1 — pronoun of class 1

POSS — possessive

PRN — pronoun

PROX — proximate

REL — relative

SG — singular number

SUBJ — subject

TA — transitive animate

TE — Bulu Salient Article

TOP — topic

^[1] I introduce the term “unpossessive” to refer to markers that are related to possessive agreement markers (e.g., are at least homonymous with, or even derived from) but are clearly non-possessive in their use.

One exception to the statement above is Alexandra Simonenko’s paper [2017], in which she proposes a detailed semantic analysis of unpossessives in Uralic and neighboring languages, focusing on poss.3sg-like markers (although she argues for a monosemic analysis). The other authors generally only compared the markers they investigated with previously described cases, appealing to common theoretical notions, such as uniqueness, familiarity, partitive specificity, etc., but without providing semantic definitions of the markers.

^[2] Unless stated otherwise, examples from the Kazym dialect of Northern Khanty are mine. When giving the context for an example, I use double quotes to indicate that the context was translated into the target language along with the target sentence. In the absence of quotes, the context was described to the consultants in the language of interaction, here, Russian.

^[3] Throughout the paper I adopt the convention to capitalize names of language-particular categories to highlight their idiosyncrasies [Haspelmath 2010: 674], e.g., writing “(Proper) Possessive” instead of “possessive” implies that the category in question must not correspond to some universal categorial type but may diverge from it in different ways.

^[4] In [Mikhailov 2021a; 2023], I called it the “Topic Marker” because I hypothesized that it marks topical referents. In what follows I will reject this hypothesis (§4.2); hence a different name is needed.

^[5] I follow the model of the Uralic family discussed in [Sámmol Ánte 2022: 3–4].

^[6] Compare Meadow Mari üdər-em-že [daughter-poss.1sg-poss.3sg] ‘my daughter (topicalized)’, adapted from [Kuznetsova 2003], where a poss.3sg in a non-possessive function attaches to an already possessed noun resulting in “possessive stacking”.

^[7] The hash sign (^#) indicates semantic or pragmatic infelicity, i.e., the marked sentence or form does not have the intended interpretation or is not appropriate in the corresponding context. The other acceptability signs used in this paper include: the asterisk (*) — ​the sentence is ungrammatical; the OK (^OK) — ​the linguist constructed a sentence (different from the consultant’s volunteered translation of the stimulus) and the consultant accepted it; and the percent sign (^%) — ​some of the consultants accepted the sentence and some rejected it.

… (^#năŋ) … means that adding the parenthesized expression (here, the Possessor pronoun) results in infelicity, and vice versa for … ^#(năŋ) …

^[8] In [Mikhailov 2021b] I have argued that the Associative Possessive is independent from the Proper Possessive.

^[9] For uniqueness, see, e.g., [Russell 1905; Link 1983; Coppock, Beaver 2015b], and for familiarity, [Christophersen 1939; Heim 1982; Coppock, Beaver 2015a], among many others, as well as the recent critical discussion in [Coppock 2022].

^[10] A discourse referent is a body of information known by the interlocutors and assumed to pertain to a single individual (although there might be no corresponding individual in the actual world) [Karttunen 1976]. A related notion is the common ground [Stalnaker 1974], the set of propositions that the interlocutors mutually believe. The discourse referents of a given context are entailed to exist by the common ground. Familiarity, thus, entails existence in the world under discussion (another commonly discussed implication of definite NPs).

^[11] The reader interested in the formal version is referred to [Roberts 2003: 310].

^[12] Since, as we saw above, the discourse referent need not be introduced by any explicit linguistic material, the term “antecedent” should be taken to refer to the corresponding discourse referent here, and not to any explicit NP.

^[13] For an account of the other putative counterexamples to uniqueness / ​familiarity, the reader is again referred to Roberts’ original paper.

^[14] Schwarz and subsequent work investigating the weak vs. strong distinction work in a Situation Semantics framework which is not directly comparable to Roberts’ dynamic-semantic framework. For a comparison see [Coppock 2022].

^[15] Arkoh and Matthewson [2013: 17–18] instead propose that weak familiarity is sufficient for both the German Strong Article and the Akan Definite Article. For our purposes, merely noting the empirical patterns presented in this subsection is enough.

^[16] In German, this sentence is rendered using the Weak Article [Schwarz 2009: 40], which corresponds with the Akan bare form according to Arkoh and Matthewson. The Strong Article is excluded here, just as expected under their analysis.

^[17] See [Serdobolskaya 2017] for a conclusion along the same lines based on the differential object marking pattern in Beserman Udmurt which involves a distinction between accusative-marked and possessive accusative-marked objects which is based on the uniqueness-familiarity distinction respectively.

^[18] See the original article for references to experimental and corpus-based work supporting this conclusion.

^[19] For an introduction to Roberts’ model of the intentional structure of discourse structured by QUDs and its applications to anaphora resolution, English Focus, and other phenomena, see [Roberts 2012; 2022].

^[20] This notion will become especially relevant when we turn to information structure in the next section.

^[21] See [Roberts 2019] for a discussion of the terminological matters and the references there. Question under discussion can be taken to refer roughly to the same notion as theme. See below for details.

^[22] Where nothing hinges on it, I adopt for Obdorsk examples the glossing conventions I use for the Kazym dialect.

^[23] For a discussion of secondary topic see below.

^[24] Sentence rheme utterances maintain Active Voice since there is no topical constituent that would need to be promoted to subject.

^[25] I adapt Lambrecht’s taxonomy of foci that Nikolaeva uses [Nikolaeva 2001: 3–4] to the terminology used here. Thus, instead of “argument focus”, I will speak of argument rheme and similarly for Lambrecht’s predicate focus and sentence focus.

^[26] Since O1 and O2 are only morphologically distinguished in verbal agreement, Nikolaeva marks the verb as either O1 or O2 depending on whether it agrees with the Object.

^[27] Here and below, I quote only the relevant parts of the hierarchies in question, following Muravyev. For details see the works cited above and the references there.

^[28] Since personal pronouns do not form Locative, a constructed form is provided here in angle brackets.

^[29] Muravyev cites [Nikolaeva, Bárány 2019] with a similar proposal that Tundra Nenets (< Uralic) possessive agreement with lexical Possessors is conditioned on their proximate status.

^[30] Bliss does not provide glosses for this example but note that anna nínsstsinaana is the same expression as the morphologically segmented ann-wa n-ínsst-innaan-wa [dem-prox 1-sister-1pl-prox], namely, ‘our sister (proximate)’. The same remarks apply to examples below.

^[31] Bliss uses the terms “topic/non-topic” for “main/peripheral character” which I don’t use here in order to avoid confusion with the information-structural topic in Nikolaeva’s sense.

^[32] The partitive article function is quite often found for 3sg possessives in Uralic languages, e.g., in Udmurt [É. Kiss, Tánczos 2018], Komi Izhem, and Meadow Mari [Simonenko 2017]. É. Kiss and Tánczos argue that the partitive article in Udmurt is not merely a function of the poss.3sg but is an independent marker. The same argument can be made for the Kazym Khanty Partitive Article, see [Mikhailov 2021a].

^[33] The non-finite form used here functions as a Mirative Past Tense.

^[34] Note that here the question under discussion (“What happened to the computer?”) ensures that the thief is in the rheme of the utterance, anticipating the discussion below that the Salient Article is not restricted to utterance theme. On the other hand, I assume that the thief is still salient in (44) by virtue of bridging. A robbery entails the existence of a thief. Since the robbery is the topic of the dialogue, the thief is salient as its agent.

Even though the computer is also salient here, it cannot be marked with the Salient Article since it is marked with a Proper Possessive which is in complementary distribution with the Salient Article.

^[35] According to my assumption from §2.3 (see [Mikhailov 2023]), waśaj-en ‘Vasya’ is marked with the Proprial Article here, and not with the Salient Article. Still, Vasya is arguably more salient than the thief, which would make him the proximate, and this excludes the Salient Article on the thief per Muravyev’s hypothesis. Exactly the same is observed with personal pronouns: the presence of a personal pronoun excludes the Salient Article on the other NP. I do not include the relevant data here for reasons of space.

^[36] Note that in all the examples based on the thief story, Associative poss.3sg is also a viable marking option for the NP referring to the thief, although, when the Salient Article is available, as in (50), it is preferred by most speakers.

^[37] Muravyev, however, does not provide any theoretical account but merely discusses preliminary observations.

^[38] For an introduction to Optimality Theory see [McCarthy 2008].

^[39] This is an oversimplification since at least proper nouns and personal pronouns also appear to compete for proximate status in Northern Khanty, even though they are not marked with the Salient Article (cf. footnote 35).

^[40] I thank Valeria Marinina (p.c.) for introducing me to Diessel’s paper.

^[41] An anonymous referee asks whether linear order affects the marking possibilities in (57). I have no evidence for the relevance of linear order, but syntactic role does have an effect here according to preliminary data. Thus, if the sentence is passivized and the girl demoted to Oblique, the preferred marking pattern is vice versa (given schematically as glosses): [boy-poss.2sg] [girl-poss.3sg-loc] [hit-pst-pass[3sg]]. I plan to investigate this further in future work.

^[42] An anonymous referee notes that this reasoning might suggest that the salience of the referent is asserted rather than presupposed. I would disagree because that would predict the Salient Article to have a far wider distribution than it in fact has. The presuppositional view instead accounts for its narrow distribution, and the reasoning here fits neatly into the general workings of presupposition accommodation, see (16), as well as §3.1 for examples from the English Definite Article.

^[43] See David Kaplan’s seminal paper [Kaplan 1989] and much subsequent work, e.g., [Kratzer 2009].

^[44] Other parameters, such as, e.g., place of the utterance, may also be added.

^[45] There are other views on the semantics of first and second persons, in particular the proposal by [Sudo 2012], which do not directly support my line of reasoning. For the lack of space, I leave the comparison of various theoretical proposals pertaining to this issue for another occasion.

About the authors

Stepan K. Mikhailov

Author for correspondence.
Email: stepanmihajlov@gmail.com
Russian Federation, Moscow

References