PRAGMATIC REASONS FOR NON-CLAUSE-INITIAL PLACEMENT OF SUBJECTS OF ACTIVE AND PASSIVE CLAUSES IN NORTHERN MANSI

. This article discusses the linear placement of the syntactic function of subject in Northern Mansi, concentrating on the less frequent non-clause-initial placement, and differences between the nature of subject in active (SA) and passive (SP) clauses. The study is realized as a template analysis. The main aim of this study is to elaborate and expand on my previous template analysis on Northern Mansi constituent order (Virtanen 2021), regarding the subjects of both active and passive clauses. The data used in the study are gathered from Лӯимā с º å рипос [L ū im ā S ē ripos], the only up-to-date Mansi newspaper. The most remarkable result of this study is that the majority of the non-clause-initial subjects are pragmatically focal, which affects the linear placement of the subject. This finding does not overrule the earlier findings concerning the correlation between primary topic and subjects, but it refines the whole picture.


Introduction
This article discusses the linear placement of the syntactic function of subject in Northern Mansi, concentrating on the less frequent non-clause-initial placement, and differences between the nature of the subject in active (SA) and passive (SP) clauses.In my data 71% of all subjects are placed in clause-initial position, but this study discusses those 29% that are not.The study is realized as a template analysis.The main aim of this study is to elaborate and expand on my previous template analysis of Northern Mansi constituent order (Virtanen 2021), regarding the subjects of both active and passive clauses.I will concentrate on those SAs and SPs that are not placed in clause-initial position and correct some inaccurate details on the motivation for non-clause-initial placement of the subject.I will also elaborate on some earlier views on Mansi information structure.

The Mansi language and its typological features
Mansi is a seriously endangered indigenous language spoken in Western Siberia and belongs to the Ob-Ugrian branch of the Finno-Ugrian language family.The areas in which Mansi has traditionally been spoken are situated in the Khanty-Mansi Autonomous Okrug of the Russian Federation.According to the latest Census of the Russian Federation (2021), there are still 1,236 native speakers possessing proficiency in Mansi, and 951 using the language actively.Practically all of them speak the Northern variant of Mansi.The number of ethnic Mansi is much greater, more than twelve thousand.The Northern variant of the language is the only one that is still used in spoken and written form.It has been provided with a standard, but due to the limited possibilities to use the language, there are deficiencies and insistencies in the standard.
Typologically Mansi is an agglutinative language, with a rich variety of inflectional and derivational suffixes.It also has postpositions and verbal preverbs.Verbal preverbs often represent grammaticalized directional adverbials, and they function as aspect markers, usually appearing adjacent to the verb.The Northern Mansi case system consists of an unmarked nominative case and five case endings: locative, lative, ablative, instrumental and translative.Unlike many Uralic languages, Mansi has no genitive case: possession and other genitive-related relations are expressed with possessive suffixes.There is only one past tense category but two active conjugation paradigms in Mansi.The subjective conjugation is used in intransitive clauses and in such transitive clauses that do not include a topical object.The objective conjugation is used in such clauses that include a topical object.The passive is used as the inverse category of the active.There are three numeral categories in Mansi: singular, dual and plural.All three numbers occur both in verb and noun inflection.

Mansi information structure
To understand the details of Mansi constituent order, one has to be aware of the basic features of Mansi information structure.As shown in my previous analysis (Virtanen 2021), there is both pragmatically caused syntactic variation, and pragmatically caused constituent order variation: pragmatic status correlates with both syntactic functions and linear placements.Constituent order is only partly dependent on information-structure factors, but some features of constituent order correlate with particular information-structure situations.In this subsection I will first present some terminological remarks in 1.2.1 and then describe the basics of information structure in Mansi in 1.2.2.

Terminological remarks
Before proceeding to the principles of information structure, some terminological remarks need to be presented.In this context, a distinction is made between syntactic functions, semantic roles and pragmatic (information-structure) roles.The analysis connects together four dimensions: syntax, semantics, pragmatics and linear order.To demonstrate the relationships between the different dimensions of the phenomenon, it is important to apply accurate terminology.The corresponding terminology is presented in this subsection.Not all the terms listed here are used in this study, but all of them are applied in the original template analysis (Virtanen 2021).Presenting a comprehensive terminology provides a clear framework for understanding the principles of Mansi constituent order.
Syntactic functions (c o n s t i t u e n t s) are referred to with well-established syntactic terms: subject, direct object, indirect object, temporal adverbial, locational adverbial, manner adverbial and agent adverbial.The category of indirect objects includes two different types: the lative-marked r e c i p i e n t and the instrumental-marked semantic p a t i e n t.The category of adverbials is simplified to include four main categories: time, location, manner and agency.In addition, there is a class of other adverbials for some less frequent cases like conditional or reason adverbials.Furthermore, the phenomenon of s c e n es e t t i n g a d v e r b i a l s is applied: a scene-setting adverbial is a temporal or locational adverb which is placed in sentence-initial position and sets a spatial or temporal framework within which the main predication holds (see e.g.Chafe 1976 : 50).Scene-setting adverbials are not a purely syntactic category, and they are discussed in detail in Section 4.2.
The semantic roles are named agent, patient, theme, recipient, locative, goal and source, respectively.For pragmatic status, I use the terms primary topic, secondary topic and focus.This threefold distinction between pragmatic functions is mainly used only in connection with the Ob-Ugrian languages (see e.g.Dalrymple, Nikolaeva 2011).It follows the same principles as, for example, Krifka (2006), but instead of a simple opposition between topic and focus, a second level of topicality is added between them.The primary topic is the most topical element in the utterance, it is the constituent the whole sentence is about.The secondary topic is also topical but less salient: while the primary topic is related to aboutness, the secondary topic is rather related to accessibility.Focus is new, unpredictable or contrastive information brought to the discussion (see e.g.Lambrecht 1994 : 118, 207;Krifka 2006;Dalrymple, Nikolaeva 2011;Virtanen 2014).In a broader sense, when analysing corpus data, we can also talk about narrative topic, which is the most topical element in the whole text or spoken entity.
Furthermore, the typological features typical for Mansi have implications for terminology.In Mansi, both active and passive conjugation paradigms correlate with the syntactic subject of the clause.I distinguish the subjects of active and passive clauses by marking them SA (Subject, Active) and SP (Subject, Passive).Secondly, we cannot ignore the fact that Mansi is a language with a high frequency of zero anaphora: a subject or a direct object is often not expressed explicitly with a nominal constituent, if it is a topical argument and referred to with a conjugation suffix.

Syntactic variation due to information structure
Mansi information structure is primarily expressed by variation between different syntactic structures (see e.g.Skribnik 2001;Virtanen 2015).Variation between the active and the passive voice is due to a correlation between syntactic and pragmatic functions: the higher an argument is in the pragmatic hierarchy (PT > ST > F1 ), the higher the syntactic function (Subject > DO > Oblique) it occupies (see e.g.Virtanen 2015).If any semantic role other than the agent (i.e.patient, recipient, sometimes even locative2 ) is promoted to the subject (i.e. it is the most topical syntactic argument), the passive is used.As É. Kiss (2019) describes, "the highest noun phrase in the Ob-Ugric sentence functions as both grammatical subject and primary topic".
In example (1) we have both active and passive clauses in one sentence, which describes the difference between active and passive very well.'The children' carry the pragmatic status of topic in every clause in this sentence -every clause is about the children -but due to the differences in semantic roles, the first clause is in the passive, the second and the third one in the active, and the fourth again in the passive.The children are the narrative topic of the whole article: the article is about their school's field trip to an outdoor museum.Their action, or what happens to them, is in focus position also in every clause of this sentence.In the first clause of the sentence, it is recounted that the children were assigned tasks to be solved (but who was assigning them is not even mentioned).In the second clause it is told that they thought about the tasks.The third clause proposes a condition: "in the event that these same children answered correctly".In the fourth clause it is told that in the event of a correct answer they were given hunting items.
(1) Н врам-ыт тит-ыгла-вē-с-ыт, тāн ном-сахт-а-с-ыт, ос ювле child-PL ask-DER-PASS-PST-3PL 3PL think-DER-PST-3PL and back мāк-ыг те лāв-с-ыт, тāн  вт акв лōмт-ыл май-вē-с-ыт right-TRANSL if say-PST-3PL 3PL bow one piece-INSTR give-PASS-PST-3PL 'The children were posed questions, they thought about them, and if they answered correctly, they were given hunting items' Besides the variation between active and passive, there is pragmatic variation between ditransitive structures (see e.g.Virtanen 2015).Northern Mansi ditransitive constructions have recently been examined by Bíró and Sípőcz (2017) from a typological point of view.Referring to the terminology of Malchukov, Haspelmath and Comrie (2010), they state that the Northern Mansi ditransitive constructions are the S e c o n d a r y O b j e c t C o n s t r u c t i o n (SOC, also called S e c u n d a t i v e A l i g n m e n t) and the I n d i r e c t O b j e c t C o n s t r u c t i o n (IOC, also called I n d i r e c t i v e A l i g n m e n t) (Bíró, Sípőcz 2017 : 44-45).Also in ditransitive clauses, both active and passive, the most topical element occupies the syntactic function of subject and is the one that the verb correlates with (Virtanen 2015).
In example (2) we have a very typical active SOC.The text is about building of new homes in an area with a pressing need for new apartments.In the previous sentences, it has been recounted how people were queuing for better accommodation, but also about building companies.In this clause, the pragmatic primary topic, and for this reason the subject, is the building companies, while the pragmatic secondary topic is the people in the queue, which occupies the syntactic function of direct object.Finally, the pragmatically focal element in the clause is the letters the companies are sending: it occupies the syntactic function of oblique and is marked with the instrumental case.
(2) [---] кол ӯнтт-ын компания-т та нти н пак-аныл house build-PCTP.PRS company-PL themselves letter-POSS.3PLте т-ыяныл (LS 1/2018) send-PL>3PL '[---] the building companies provide them [the people in queue] with letters themselves' If we have a closer look at example (1), it includes a passive SOC in the fourth clause of the sentence: (3) [т]āн  вт акв лōмт-ыл май-вē-сы-т 3PL bow one piece-INSTR give-PASS-PST-3PL 'They were given hunting items' In this case, a passive construction is chosen, because the most topical element of the clause, 'they [the children]', is not the semantic agent of the clause.Further, the the pragmatic focus is represented by the hunting items provided.Similarly to the active, also in a passive SOC the pragmatic focus is in the function of oblique and marked with the instrumental case.
Example (4) represents a very typical IOC.In the first two clauses of this sentence, it has been told how there are houses built, and how the commune buys the houses.In this clause it is described how the already mentioned houses are given to citizens by the commune.The first person plural pronoun is the primary topic of the clause, because the speaker describes what they are doing themselves, so it occupies the function of subject.The houses that are referred to with an objective-conjugation ending are the secondary topic of the clause, so they occupy the function of direct object.Finally, the focus of the clause, the new information provided, is the people who are provided with the houses.
(4) [---] мā н тыт миркол пā л-ыл мā хман-ув-н очередь 1PL here commune side-INSTR people-POSS.1PL-LATqueue щирыл мыг-анув (LS 1/2018) according.togive-PL>1PL 'We on behalf of the commune, give them[the apartments] to our people according to the queue' In addition to variation between different syntactic structures, there is also pragmatically caused variation in constituent order: the most topical element tends to occupy the clause-initial position, while the focus of the clause is placed in preverbal position (see Virtanen 2021).Some marginal features connected to this principle are discussed in this article.

Research data and methodology
In this section, I will first present my research data in Section 2.1.Template analysis as a device is described in 2.2, and constituent order as a typological question in 2.3.

Research data
In this study, I use the same data that was gathered, classified and analysed for my previous analysis.My data are gathered from Лӯимā сº э рипос [Lūimā Sēripos], the only up-to-date Mansi newspaper, published twice a month in Khanty-Mansiysk.An average issue consists of 15-20 pages.Although the data is from one single source, the genres published within it vary, e.g. standard news articles, interviews, letters from readers and folklore texts.
This source was chosen mainly due to its contemporary nature and sufficient variety.The sources for written Mansi are limited, because the language is used only in restricted circumstances.The easiest way to include folklore texts was to gather them from the same source as all other data, because the same Cyrillic orthography is applied.The data naturally include various sentence types, such as declaratives, questions and imperatives, which vary according to the genre of the text.When something in the analysis is connected to the sentence type, this is clearly stated.Most often the same model can be applied to any sentence type, as the next section will show.
Altogether 12 articles or other texts are included from the 2014, 2017 and 2018 years of the newspaper, a total of 676 entries.The 12 articles were chosen so that different genres, variable topics, and different authors are represented, but also so that the length of an individual article is between 20 and 150 clausal entries.Among these 676 entries, 115 of them include a non-clause-final SA or SP.

Template analysis as a device for describing linear order
My previous study (Virtanen 2021) was realized as a template analysis, and this current study is mainly just a more detailed elaboration of the earlier analysis.As mentioned above, Northern Mansi constituent order is dependent, on the one hand, on syntactic functions, and on the other, on the order of pragmatic status.In this and the previous study, a two-level (syntax + pragmatics) template analysis is used to describe this complex two-level system.The implementation of my previous analysis is described in detail in Virtanen (2021 : 207-208), and the further details concerning the implementation of this study are presented in Section 4. In this section, I will describe the general principles of a template analysis.
The term template includes various kinds of analyses of different levels of language structure and is not meant to refer to a strictly delimited device.
A template can be used for describing morphological, morphophonological or -as is done here -syntactic linearizations.There is no single correct way of doing template analyses.As Good humorously expressed it, templates are "a twice incoherent class of phenomena" (Good 2016 : 22) or a "wastebasket" (Good 2016 : 27).Though many kinds of approaches are taken, a common feature is that the final result is a linear representation of the components involved.
Good (2016 : 40-103) makes an effort to provide a settled template terminology, which he calls a description language for templates.A template analysis includes four basic phenomena: stricture, foundation, desmeme and component (Good 2016 : 53-54).Component is an immediate subconstituent of a given templatic construction; in this study, components are syntactic (phrasal) constituents that occupy numbered slots inside the template.Stricture is a feature that classifies the nature of the linearization specifications in a given template (length or order) (Good 2016 : 66).In this study, the stricture is order: the main question concerns the linear order of the syntactic constituents (components).Foundation is a feature describing how components of a template are organized into an overall templatic form; the foundation can be either span or arch type.A span foundation includes left-support components and right-support components, and the remaining components (Restkomponenten) are placed between them.An arch foundation is built around a keystone, which is the topmost component at the centre (head) of the template.My analysis represents the span model with a right-support verb and a left-support primary topic: both syntactic and pragmatic functions are involved simultaneously.Finally, desmeme is used as a synonym for template: it refers to the whole linear pattern analysed (Good 2016 : 65).The desmeme is here represented by a clause.
Before proceeding, it is worth noting that this template analysis does not exactly adhere to any kind of "Good's model", although it can be described within the framework of Good's terminology.For the needs of this study, in the light of recent research and the nature of my data, the template analysis had to be applied to the pragmatic level.This kind of template analysis enables accurate and interdimensional observation of the linear order of constituents when the order is affected by both syntactic and pragmatic factors.This model results in partly very similar descriptions as Rizzi (1997) or É. Kiss (1995).Rizzi and É. Kiss, however, focus on realization of the main pragmatic functions, while my template analysis gives an interdimensional description of the affects of both pragmatic and syntactic factors.

Constituent order as a typological question
Word order can be examined from several perspectives, which are situationdependent.In his word-order handbook, Song (2012 : 3-4) sees four approaches as relevant to linguistics today: linguistic typology, generative grammar, optimality theory and performance-based theories.Among those, the approach in this study fits best the framework of performance-based theories: as for example Hawkins (1994) says, language orders its constituents in such a way that a listener can identify the scope of each constituent as quickly as possible.However, while Song intends to be comprehensive, he wholly ignores information-based theories: he only briefly refers to some approaches claiming that they do not play a remarkable role in word-order-related research, and they fail to address the issue of grammaticalized word orders and correlations (Song 2012 : 6-7).Contesting these views, the pragmatic or information-structure level is essential to the analysis in this study.
The starting point of both this study and my previous analysis (Virtanen 2021) is the question of whether Mansi constituent order is conditioned by syntactic functions or pragmatic status.The difference between languages with fixed and pragmatically conditioned word order can be described using the opposition of configurational vs. non-configurational languages (see e.g.Hale 1983; Baker 2003).In configurational languages (for example English), syntactic functions occupy fixed places within the sentence structure, while in non-configurational languages the placement of a single constituent is dependent on other factors.É. Kiss (1995) describes Hungarian -a close relative to Mansi -with the term discourse configurational language: in a discourse configurational language, pragmatic functions occupy fixed places within the syntax.Similarly Vilkuna (1989;1995), has described Finnish word order as a discourse conditioned system.
As shown in Virtanen 2021, Mansi is neither a purely configurational nor a non-configurational language: some syntactic functions have fixed positions inside the syntax while others are placed according to their pragmatic functions.Van der Wal (2009) has come up with a similar conception: languages are not purely configurational or non-configurational, but rather they lie on a continuum of different features.Van der Wal (2009 : 134) also questions the whole phenomenon of non-configurational languages.Following Van der Wal's views and taking into account my own preliminary observations on Mansi constituent order, I ended up realizing my previous analysis (Virtanen 2021) in the form of a complex two-level template approach.A template approach is an ideal way to observe the variation in constituent order from several perspectives: this way a partly configurational or discourse configurational language can be described on both the syntactic and pragmatic levels simultaneously (see Section 3).This shows that what we often call variation, actually only varies in terms of syntactic functions, because from the interdimensional point of view, constituent order is very consistent.The principles of template analysis are presented in Section 2.2, and the key results of my previous analysis in Section 3.

Template analysis on Mansi constituent order: Virtanen 2021
In this section, I will present my previous study, which is expanded in this study.I have examined Northern Mansi constituent order and its pragmatic variation in a template analysis (Virtanen 2021), which proved that the arrangement of constituent order is twofold.First, the basic constituent order is connected to the order of syntactic functions.Secondly, particular syntactic functions alternate between two or more slots, depending on which pragmatic status they represent.In other words, the pragmatic function of the constituents overrules their syntactic function.One main argument for my analysis is that the same template model can be applied to both active and passive clauses (Virtanen 2021 : 228.)The analysis (Virtanen 2021) was realized by dividing the data into clausal entries and providing desmemes (chains of constituents) out of clauses.A table with desmemes on horizontal lines was created, so that the components (constituents) of each desmeme could be adjusted with the slots (vertical columns).The same analysis was completed on both syntactic and pragmatic levels: during the syntactic phase, the slots were named with syntactic functions, and during the pragmatic phase with the names of pragmatic status.Finally, the two analyses were brought together in one template, and the template was turned to a vertical position.
As a result of the above-mentioned analysis, I presented the template model of 9 + 1 slots shown in Table 1 (see also Virtanen 2021 : 228).On the right side of the template, we can see the default placements of syntactic functions.On the left side, the placements of pragmatic status are shown.Any syntactic function, except predicate, can be placed in the pragmatically marked slots due to their focality or topicality -in any other case they are placed in their default slots.The predicate is placed in clause-final position without exception.

Table 1
Placement of syntactic and pragmatic functions in the 9 + 1 slot model (Virtanen 2021 : 228) *) An empty cell means that the pragmatic or syntactic perspective is not relevant to the slot in question.
A more detailed explanation and justification for the solutions in this model are presented in Virtanen (2021).Just to mention briefly, there are neutral fixed default placements for adverbials (Slots 2-4), and Slot 5 for additional adverbials following them.Slot 6, termed D-Slot in my analysis, is for direct and indirect objects and directional adverbials, in the event that they are pragmatically neutral or secondary topics.It is not surprising that these syntactic functions share a common slot, seeing that the semantic roles they represent are also close to each other.Slots 1 and 8 are pragmatically motivated: Slot 1 can be occupied by only a pragmatically topical constituent, and Slot 8 only by a pragmatically focal constituent.In addition, there is Slot 0 for scene-setting adverbials, which carry pragmatically complicated features, and may be followed by a topical constituent.Placement of preverbs and conjunctions is not discussed in my analysis (Virtanen 2021), but statistics on their appearance in different placements is shown in the appendix of the article.
As shown in the template above, the syntactic function of subject can occupy two alternate positions.The clause-initial default placement of the subject is due to the correlation between syntactic functions and pragmatic status (the pragmatic primary topic occupies the syntactic function of subject), and the tendency for primary topic to occupy clause-initial position.74% of SAs and 63% of SPs are placed in clause-initial position due to their topicality.Regarding SAs and SPs as a complex group of subjects, the corresponding share is 71%.Furthermore, according to the aforementioned analysis (Virtanen 2021), those subjects accompanied by a scene-setting adverbial are placed in non-clause-initial position, just after the scene-setting adverbial.However, on closer inspection, this has proven to give only a partial explanation: it is not incorrect, but it does not explain the non-clause-initial placement in all cases.In this article, I will expand on the description and present further reasons (beside the presence of a scene-setter) for non-clauseinitial placement.I will also discuss the nature and definition of scene-setting adverbials.

Focusing on non-clause-initial subjects
In this section, I will first briefly discuss the main aims of this study, and how this study expands on the previous analysis, in 4.1.Scene-setting adverbials and some complex questions on them are presented in 4.2.Finally, the implementation of this current analysis is described in 4.3.

Open questions on non-clause-initial subjects
Implementation of the more extensive template analysis of Mansi constituent order is described in detail in Virtanen 2021 : 207-208.The aim of the study was to realize a detailed template analysis of the linear order of constituents on both syntactic and pragmatic levels, and to combine the results into a complex linear model.The preliminary data classification and the actual analysis were realized in Excel in two phases by simple adjusting (in the first phase) syntactic functions and (in the second one) pragmatic status with vertical columns according to their regularly repeating linear placements.
As a result of comparing the syntactic and pragmatic analysis, the table in Section 3 was presented.
According to the previous analysis (Virtanen 2021), there are two possible linear placements for the syntactic function of subject.The same model is applicable to both active and passive clauses.The default placement for subject is the clause-initial slot, and it is justified with pragmatic factors, i.e. the high topicality of the subject (see Section 2.2).Still, 26% of SAs and 37% of SPs are placed in non-clause-initial position.As mentioned in the conclusion section of my previous study (Virtanen 2021), there are still unsolved questions concerning non-clause-initial subjects.This is mainly due to inaccurate pragmatic analysis, which requires some closer inspection.
As described in my previous study (Virtanen 2021), the pragmatic primary topic tends to occupy the clause-initial position.Due to the strong correla tion between the pragmatic primary topic and the syntactic function of subject, the syntactic subject very often occupies the clause-initial position.As mentioned above, only an adverbial can occupy the clause-initial position instead of the subject.Pragmatically neutral adverbials have a fixed order of positions inside the syntax (time -place -manner), but a scenesetting adverbial occupies the absolute clause-initial position.The linear order with a clause-initial adverbial could be justified with the following explanations: 1) an adverbial occupies the clause-initial position due to its topicality, 2) a subject occupies the non-clause-initial position due to its focality, 3) an adverbial lacks topicality but occupies the clause-initial position due to its semantic scene-setting function.Which one or which ones are relevant based on my corpus data?
To answer this question, a new pragmatic analysis of non-clause-initial subjects has been carried out to determine in greater detail what causes the non-clause-initial placement.Concentrating exclusively on the non-clause-initial subjects, their pragmatic roles, and the possible existence of scene-setting adverbials, a more detailed description has been created.The syntactic level of the earlier analysis will also be slightly elaborated upon, but bigger changes are not needed.The statistical data presented in the following sections is based on the elaborated, most recent analysis.

Features and nature of scene-setting adverbials
Scene-setting adverbials represent a specific category including several types of adverbials and carrying a particular semantic and pragmatic status.A scene-setting adverbial is a temporal or locational adverb which is placed in sentence-initial position and sets a spatial or temporal framework within which the main predication holds (see e.g.Chafe 1976 : 50).As mentioned in Virtanen 2021, in my data scene-setting adverbials share the sentenceinitial position with the topic.Regarding the information-structure principles of Mansi, it is appropriate to pay attention to the possible topicality of scene-setting adverbials.The question has been touched on and discussed in the literature.For example, Benincà and Poletto ( 2004) state that scenesetting adverbials are placed within the topic field.Furthermore, for example Rizzi ( 2004) has stated that they are not part of the topic, as they do not correspond to the definition of topic, but rather they occupy an independent position.
The nature and definition of scene-setting adverbials play a key role in this study.Do the semantic features of clause-initial adverbials affect the non-clause-initial placement of subjects, or is the clause-initial placement of adverbials due to pragmatic features of subjects?How can we distinguish scene-setting adverbials from other kinds of clause-initial adverbials: how is the difference defined?Or are there any other kinds of clause-initial adverbials in Mansi?In my previous analysis (Virtanen 2021), I classified all clauseinitial adverbials as scene-setting adverbials.Moreover, it was seen in my data that at least some of the clause-initial adverbials shared similar features with the pragmatic status of topic.In my data, almost all clauses with a non-clause-initial subject include a clause-initial adverbial.In other words, the clause-initial slot is occupied by either a subject or an adverbial, only exceptionally by some other syntactic function.Should all clause-initial adverbials be considered as scene-setting adverbials, and do they explain the whole question of non-clause-initial subjects?
There are three questions to be taken into account.First, what is the exact definition of a scene-setting adverbial?Secondly, what is the pragmatic status of the other syntactic functions in the same clause?Thirdly, how can we distinguish between scene-setting adverbials and other possible kinds of clause-initial adverbials?As can be seen above, the given description does not include any information on the exact pragmatic status of scenesetting adverbial, but it is still a comprehensive definition: a scene-setting adverbial, unlike a participant related modifier, sets the framework for the whole predication.Typologically, a clause-initial adverbial is not automatically a scene-setting adverbial, but as shown in my data, in Mansi scenesetting adverbials are in practice the only adverbials that occur in the clauseinitial position.Furthermore, the presence of a focal subject does not exclude the possibility of the existence of a scene-setting adverbial: the existence of both a focal non-clause-initial subject and a clause-initial scene-setting adverbial is wholly possible.
In the following sections, I also intend to outline the nature of scenesetters by elaborating on my corpus data.I will also demonstrate how the nature and motivation of an individual non-clause-initial subject can be classified as one of the categories (A-C) mentioned in the following section.

Implementation of the analysis
As described in Section 4.1, this study represents a continuation of the analysis in Virtanen (2021), and the same, already classified data is used.The data used in this study is restricted to those clauses that, in my earlier analysis, were classified as including a non-clause-initial subject.That data was in the present study further classified according to the pragmatic status, i.e. the topicality or focality, of the subject.Moreover, also the clause-initial adverbials were analysed in the same way.All of the 115 non-clause-initial SAs and SPs, and adverbials preceding them in my data are once more given labels referring to pragmatic status.The pragmatic status of individual referents is defined according to their appearance in the whole text, in the same paragraph, in previous sentences and in the immediately previous sentence.Inherent topicality is also taken into account.
After the more precise pragmatic analysis, some rearrangements have been made in the organization of the original syntactic analysis.The syntactic analysis is elaborated upon by making the following changes: A. Subjects that are found to be focal in the pragmatic analysis are moved to the focus slot (Slot 8) if possible.In practice this means relocating subjects from Slot 7 to Slot 8, whenever Slot 8 is free.B. Clause-initial adverbials that are found to be pragmatically topical, or are not placed in their default slots, are moved to Slot 0 (the scene-setter slot) if possible.In practice, this means that adverbials are transferred from Slots 2, 3 or 4 to Slot 0, whenever slots from 0 to the one in question are free.C. Subjects that are found to be grammatically topical and placed in Slot 7 are moved to Slot 1 if possible.This is possible in the event that there are no other arguments placed in slots between Slot 1 and Slot 7.
As a conclusion of the focused analysis, 115 clauses with a non-clauseinitial subject are divided into three groups: 1. Clauses including a clause-initial adverbial in Slot 0, a topical subject in Slot 1, and other arguments following the subject in their default slots.2. Clauses including adverbials all in their default slots, and a focal subject in Slot 8. 3. Clauses including a clause-initial adverbial, possibly other arguments following it, a subject in Slot 7, and a focal argument in Slot 8.
Examples and further elaboration are provided in Section 5.

Motivation for non-clause-initial placement of subjects in my data
Unlike my previous analysis (Virtanen 2021), this more detailed analysis shows that there are more complicated reasons for the non-clause-initial placement of subject than the existence of a scene-setting adverbial.As a conclusion of the above-described analysis, the non-clause-initial subjects can be divided into three different groups (shown above and elaborated upon in the following subsections).As the main result of this study, I can state that the majority of the SPs (78%) and more than half of the SAs (52%) placed in non-clause-initial position are pragmatically focal.Focal subjects are not typical or frequent in Mansi, but in marginal situations they are possible, especially in passive clauses, where the possible impersonality (no agent) of the clause triggers a situation with the pragmatic focus occupying the syntactic function of subject.In this section, I will first discuss those clauses with a topical subject accompanied by a clause-initial adverbial in 5.1.Further, focal subjects accompanied by a clause-initial adverbial are discussed in 5.2, and some challenging cases found to lie in between the above-mentioned types are gathered in 5.3.Finally, a comparison between active and passive clauses is presented in 5.4.

Non-clause-initial topical subject accompanied by a clause-initial adverbial
In 33% of all adverbial-initial clauses, the syntactic subject is in the template model placed in Slot 1, the default Subject Slot.Moreover, 39% of all nonclause-initial subjects can be regarded as pragmatically topical.In other words, in these clauses the syntactic subject represents a notion that has been mentioned in the immediately preceding sentences or has appeared several times in the preceding discussion.Also inherently topical notions are taken into account.
At the same time, the clause-initial adverbial functions as a scene-setter for the whole action but does not take the status of topic over from the subject.In this subsection, I focus on those examples where the subject is both pragmatically topical and placed in Slot 1.
In example ( 5) we have a topical subject: the people living in containers have been mentioned in several previous sentences.The pragmatic focus of the clause is the houses the families are buying: it is the new information provided.The clause-initial slot is occupied by a temporal adverbial.By combining the pragmatic and the syntactic analysis together, we can see that the temporal adverbial is not taking over the pragmatic status of topic but setting the temporal scene for the whole clause.In the same way in (6), we have a very clear narrative topic: the person referred to with a third person personal pronoun has been under discussion in the previous sentences and in the whole text.In the previous sentence it was told what he did in the mentioned year, and now it is told how he proceeded to theatre work in the same year.Again, the clause-initial temporal adverbial is not taking over the pragmatic status of topic but setting the temporal scene for the clause, while the subject occupies its default slot before all other constituents, which are also in their default slots.7) includes a pragmatically focal direct object.The temporal adverbial in clause-initial position functions as a scene-setter.The man handing the belt over to the president is the subject of the clause and pragmatically topical: he has been mentioned in several previous sentences.The belt is now mentioned for the very first time, and new information provided is that the president was given a traditional male belt of the Mansi people.As presented in my previous studies (Virtanen 2013;2015), the focal direct object is not morphologically marked.

Non-clause-initial focal subject accompanied by a clause-initial adverbial
42% of all non-clause-initial subjects are placed in Slot 8, the default Focus Slot, and in 58% of all adverbial-initial clauses the syntactic subject can be regarded as pragmatically focal.In these cases, the subject represents a notion that has not been mentioned in the recent discussion/text, or which is unpredictable in the context.In most of these cases, the clause-initial adverbial carries a very similar scene-setting function as in the clauses mentioned in 5.1.In other words, the function of the clause-initial adverbial is trivially scene-setting, but at the same time the subject of the clause is genuinely focal.In this subsection, only those clauses are discussed where the subject is both pragmatically focal and placed in Slot 8.
In examples (8), ( 9) and (10) the subjects of the clauses are placed in pre-verbal positions, i.e. the Focus Slot in my pragmatic analysis.In (8), the new information provided in the clause is that there are reindeer shepherds.They have not been mentioned earlier, and now their existence in the area is provided as new information.So, the syntactic subject is pragmatically focal.At the same time, we cannot call the clause-initial local adverbial a topic, because it has not been mentioned in the previous sentences, and there is no connection that could make it topical.As a conclusion, we can say that the clause includes both a scene-setting adverbial in Slot 0, and a focal subject in Slot 8. Example ( 9) is from the beginning of an article about the "Freaky Street" New Year's program in Khanty-Mansiysk.In the third sentence of the first paragraph, it is told that traditional chums were installed along a riverbank.The chums are the syntactic subject of the clause, but they are also pragmatically focal, and they occupy the Focus Slot in the template analysis.Further, there is a local adverbial in clause-initial position, which is not topical according to the pragmatic analysis, but can be regarded as a scene-setting adverbial.

(9) Нā лми вā та-т āнох-ныл вā р-им мā хар-ыт ляпат
Nalmi shore-LOC bronze-ABL make-PCTC.PRET mammoth-PL beside онтолов  рнкол тӯщтал-ав-ес (LS 1/2018) nine chum raise-PASS-PRT.3SG'Nine chums were installed along the Nalin riverbank alongside bronze sculptures of mammoths' A couple of sentences later, it is explained that there are real indigenous people working in the chums.Now the chums can be regarded as topical elements, because they have been mentioned a couple of sentences earlier, and the sentence brings information about them.The new information is that there are indigenous people working in them.The high topicality of the clause-initial adverbial in (10) raises the question of whether the local adverbial should be defined as a scene-setter or the topic of the clause.This is one of the rare examples in my data where the locative adverbial is a clearly topical argument (but in many other cases it is not).Is it possible to distinguish between scene-setting adverbials and topical adverbials, and if so, how can we define the difference?Furthermore, do these definitions exclude each other or not?These questions will not be answered here, but it is good to be aware of the pragmatic complexity of the clause-initial adverbials.

Non-clause-initial subject accompanied by a clause-initial adverbial and a pragmatically focal argument
In this subsection, such clauses are presented where the non-clause-initial subject fits neither the requirements of the topical subjects presented in Section 5.1, nor the focal ones in 5.2.18% of all non-clause-initial subjects are placed neither in the default slot (Slot 1), nor in the pre-verbal slot (Slot 8), but in Slot 7, which was the slot for non-clause-initial subjects in my original analysis.Further, some of these subjects are found focal in the pragmatic analysis, while some others are found topical.Whenever the syntactic subject is in Slot 7, there is another constituent, often a pragmatically focal one, occupying Slot 8, and also another (possibly topical) constituent occupying the clause-initial position.In 3% of all clauses with a non-clause-initial subject, the subject is regarded as pragmatically neutral.
In both examples ( 11) and ( 12) the subject of the clause is a personal pronoun, and there is a pragmatically focal adverbial placed in Slot 8.In addition,in (11), there is a clause-initial temporal adverbial followed by a local adverbial, while in (12), the clause-initial constituent is a topical direct object.In both cases, the first person subject in Slot 7 cannot be defined as pragmatically focal.Example (11) is from an article about health issues of the local people.A local Commissioner for Human Rights talks about the situation in Yugra.In this clause she mentions herself for the first time in her speech, but due to the context (talking about her own work), the fact that first person personal pronouns are usually inherently topical, and the nature of the other constituents in the clause, the subject of this clause cannot be regarded as focal.The clause-initial adverbial is a typical scene-setting adverbial, and the local adverbial in pre-verbal position is defined as a focus due to its lacking an appearance in the previous sentences.Example ( 12) is from a text about the residential construction companies working in the region.The companies are mentioned in the previous sentences and in the first clause of this very sentence: in this second clause the representatives of the company are referred to with a personal pronoun.In the pragmatic analysis, the third person pronoun is defined as a secondary topic.The focus of the clause is the manner adverbial describing the level of familiarity of the people.11) and ( 12) clearly represent situations where the subject is not focal but cannot be placed in its default position either.The scene-setting adverbial in ( 11) is followed by another adverbial in its default position.The direct object in ( 12) is placed in clause-initial position preceding the pragmatically topical subject, similarly to scene-setting adverbials.In other words, despite the topicality of the subjects, the examples in this section cannot be defined as belonging to the same category as those in Section 5.1.
Finally, many of the constituents placed in Slot 8, following a subject in Slot 7, are such arguments that can be regarded as parts of the predicate: negation particles, preverbs, infinitive or participle forms of verbs, or conjunctions.As a consequence of this remark, the analysis could be modified so that all pre-verbal elements -infinitives, participles, negation particles and preverbs -are included in predicates in Slot 8, and in this way regarded as belonging to predicates.By this modification we could move most of the focal subjects in Slot 7 to Slot 8.
Example ( 13) is from a text about the poor circumstances in the local countryside.Exceptionally the whole two-clause sentence is presented here.The sentence includes a clause-initial local adverbial, which can be defined as a scene-setting adverbial, two clauses each with a subject ('electricity' and 'gas', respectively) and a predicating negative particle ('there is not') in each of the two clauses.According to the pragmatic analysis, both subjects are focal.The previous sentences, and the whole text so far, are about the life of an old man.His education and work career have been discussed in the previous sentences, and in the immediately preceding sentence it was mentioned that he stayed in the same village until he reached old age.At this point it is also stated that the village has no gas or electricity.The state of the village's infrastructure has not been discussed and is not an expected topic in this context.
(13) Та пā в-ылт электричество ā тим, газ ā тим (LS 9/2014) that village-LOC electricity is.not gas is.not 'In that village, there is no electricity and no gas' Examples ( 14) and ( 15) are two sequential sentences from the same text.Both of them include an infinitive form of the same verb.Both of them include also a nominal subject which is found to be focal in the pragmatic analysis.The subject is not placed in the focus slot, but if we regard the complex verb forms with an infinitive and a conjugated verb as predicates, we end up with a situation where the subject is in Slot 8 and the predicate in a normal way in Slot 9.However, in ( 14), there is also a local adverbial placed between the infinitive and the predicate verb, which does not support the option of moving the subject to Slot 8 and including the infinitive in the predicate.Only if also the adverbial is regarded as a part of the predicate, can we reach a model where the pragmatically focal subject is in the default Focus Slot.

Non-clause-initial subjects of active and passive clauses in my data
So far in this article, a distinction has not been made between the active and passive clauses, even though both verb moods have been presented alongside one another.Here I shall provide very briefly the most important statistical differences and some consequences arising from them.Before proceeding to statistics on different non-clause-initial subjects, it is good to recall that only 21% of all subjects belong to this group.In other words, this article discusses a marginal type of subjects, and the following statistics concern only this marginal group.Clause-initial subjects are not included in this study.
Regarding non-clause-initial subjects placed in the default slot of subject, preceded by a clause-initial adverbial in Slot 0, the active sentences represent a clear majority.91% of non-clause-initial subjects in Slot 1 are SAs, whilst 85% of the subjects in Slot 7 and 68% of subjects in Slot 8 are SAs.However, these numbers do not tell the proportional share of the placements.Only 21% of all non-clause-initial subjects are SPs, so we need another perspective to see the real shares.38% of all non-clause-initial SAs and 13% of non-clause-initial SPs are placed in Slot 1 following a scene-setting adverbial in Slot 0. In Slot 7, the corresponding numbers are 26% of all non-clause-initial SAs and 22% of all non-clause-initial SPs.Finally, 36% of all non-clause-initial SAs and 65% of non-clause-initial SPs are in Slot 8.
These numbers prove that there are relatively more topical non-clause-initial SAs than SPs, but still the focality of the subject is the most frequent reason -or at least one factor -for placing an individual subject in non-clause-initial position.SPs are pragmatically focal more often than SAs, which is in connection with the possible impersonality of passive clauses: not all passive clauses are impersonal -many of them include a lative-marked agent adverbialbut in those ones that do not include an agent adverbial, the syntactic function of subject is often occupied by the focus of the clause.Also active clauses include focal subjects, just in smaller proportion.

Conclusions
This study did not bring any large-scale changes to my previous template model (Virtanen 2021), but the motivation for placing a subject in non-clause-initial position was elaborated upon in a more accurate way.The differences between subgroups of non-clause-initial subjects are now described in detail.These kinds of differences can be observed only by combining the pragmatic and syntactic analyses: when the syntactic analysis is expanded by relocating constituents inside the template according to their pragmatic status, we can find systematic occurrences in the syntactic level as well.Still, this causes no modifications to the template presented in Section 3. The main results of this analysis can be summarized as follows: 1.The most frequent reason for placing an SP in non-clause-initial position is the focality of the syntactic subject, in connection with impersonality of the passive clause (no agent).65% of all non-clause-initial SPs are placed in the Focus Slot, and there are also pragmatically focal SPs placed in Slot 7. Altogether 78% of non-clause-initial SPs are regarded as pragmatic foci.2. Also more than half (52%) of the non-clause-initial SAs are pragmatically focal.36% of all non-clause-initial SAs occupy the default Focus Slot.In addition, some SAs placed in Slot 7 are pragmatically focal.3. 46% of the non-clause-initial SAs hold topical pragmatic status.38% of all non-clause-initial SAs are placed in the subject's default slot (Slot 1), and only a scene-setting adverbial in the absolute clause-initial position causes them to take a non-clause-initial position.Further, there are topical SAs placed in Slot 7.
In my previous analysis the non-clause-initial placement was justified with the presence of a scene-setting adverbial.The most remarkable information provided in this study is that the majority of the non-clause-initial subjects are focal, which affects the linear placement of the subject.The existence of focal subjects is a marginal phenomenon on a large scale, and in the light of my previous studies (Virtanen 2014;2015), not a default option, but still a linguistically justified situation.This finding does not overrule my earlier findings concerning the correlation between primary topic and subjects, but it refines the overall picture.
There are minor differences between the motivation for placing SAs and SPs in non-clause-initial position.There are relatively more focal subjects in passive clauses than in active clauses.Active clauses with a subject in the default place and only a scene-setting adverbial preceding it are overtly and relatively more frequent than passive clauses.
Furthermore, the question of topicality of a scene-setter has been discussed in the literature (Rizzi 2004;Benincà, Poletto 2004) and is not easy to solve.According to my data, the simultaneous existence of a scene-setting adverbial and a focal subject is fully possible, but it does not automatically make the scene-setter a pragmatic topic.In this article I have, in some contexts, used the term "clause-initial adverbial" to make it clear that I am not expecting all clause-initial adverbials to be scene-setting adverbials.It is shown in my analysis that in most of the cases, the clause-initial adverbial is actually a scene-setting adverbial.Still, this is another question that demands further elaboration.
The results of this study have an effect also on my earlier studies on Mansi information structure.In Virtanen (2014; 2015) I have described the strong correlation between the syntactic function of subject and the pragmatic status of primary topic.As shown in this current study, also focal subjects are possible, especially in passive impersonal clauses.Still, this finding does not overrule my earlier observations: discovering the existence of pragmatically focal subjects brings clarification to a marginal phenomenon, which has no effect on the fact that in the great majority cases, the syntactic subject is the pragmatic primary topic of the clause.