ON DATIVE-LATIVE ENCODED DIRECT OBJECTS IN WEST MANSI

. Among the varieties of Ob-Ugric, West Mansi stands out in showing a differential object marking pattern in which direct objects may be encoded with a multi-functional dative-lative case. The present study builds on the West Mansi texts in the Ob-Babel corpus, where approximately one third of all referential and given objects is marked. As has been observed for other object marking varieties of Mansi too, the majority of marked objects agree with the verb. Object agreement signals topicality, but in focal contexts given objects also do not agree. A single parameter responsible for object marking could not be identified, and the results point to a bundle of parameters falling under the notions of prominence and activation. Gram-maticalization of the dative-lative case as a marker of direct objects may result from topicalization, recipient promotion, and agent marking. The study contributes to the knowledge about DOM in Mansi, and adds West Mansi to the body of literature on accusative-dative syncretic DOM.

(1) a. Via passive construction promoted goal subject: taew kuttəxtalt-əs, jox w talt-əs, yø mp-t-nə porr˝m-ow-s kin-i:-now daughter-INS look_for-PRS-SG<1PL 'We'll find you (~ we'll equip you with) an even better girl of princely origin' (P 1268.69)Indexing a topical DO's number on the verb (differential object agreement, DOA) is a grammatical trait across all Ob-Ugric varieties.DO marking (or flagging) applies in general only to personal pronouns, which show a distinct oblique form (1b) or even a special accusative form in all varieties of Ob-Ugric (Skribnik, Laakso 2022 : 528, 530).Lexical nouns and other pronouns show either no object marking at all, as it is the case in Khanty and North Mansi, or there is asymmetric differential object marking (DOM) with two subtypes (cf.Lavotha 19532 ): (i) zero marking vs a c c u s a t i v e f l a g g i n g in East Mansi (Konda) (Wickman 1955 : 65-71;Virtanen 2015), South Mansi (Tavda) (Honti 1969;Sherwood 1996;Szilágyi 2014), and parts of West Mansi (Middle and Upper Lozva) (Munkácsi 1894 : 105, 156-157); (ii) zero marking vs d a t i v e -l a t i v e (DLAT) f l a g g i n g in the West Mansi varieties of Pelymka and North Vagilsk (Marcantonio 1993).The main object marking parameters which have been identified for Mansi are prominence of the DO referent (pragmatic definiteness) as well as semi-activeness in the discourse (see Section 2).In Middle Lozva, in addition to the accusative, the dative-lative also seems to occur on direct objects.Middle Lozva is thus a candidate for a third type of Mansi DOM, in which asymmetric DOM (zero vs case-marked DO) would co-occur with symmetric DOM (accusative vs dative-lative).The scarce data, however, is hardly sufficient for a robust establishment of such a type (see Section 4).
The designation dative-lative is due to the dative function with animate referents and the lative function with inanimate ones.In general, the Mansi dative-lative is a multi-functional case, see Figure 1 and examples (2a-2f).Four functions are cross-dialectal: indirect object and goal marking, agent marking in passive and participial clauses, and formation of depictive adverbials. 4 A fifth function is characteristic for all of West Mansi, namely separative (ablative) marking due to a merger of the Mansi ablative case with the dative-lative.

309
3 Lavotha follows his teacher Miklós Zsirai, who had expressed this idea in his university lectures (Lavotha 1953 : 210).In fact, the possessor suffix of the 1SG has also been named as the source of the Proto-Uralic accusative (see Honti 2022 : 10-24 for a recent discussion). 4An anonymous reviewer considers the possibility that the marker of depictive adverbials -nə is a derivative suffix rather than the dative-lative case.An answer to this question is beyond the scope of this paper.Depictive adverbials are mentioned in 3.5 with the aim of excluding false direct object readings.For the analysis of DOM in Pelymka and North Vagilsk, the etymology of the depictive marker is of minor importance.

NV & P (& ML)
-direct object (2f) In an active clause, a DLAT-encoded DO may occur together with a DLATencoded goal expression, and both may depend on the same verb (3a).In a passive clause, a DLAT-encoded agent may co-occur with a goal expression (3b).5Note that DO and agent marking are in complementary distribution: the first occurs only in active clauses, the other one only in passive clauses.
(3) a. ass o:t˝r pyw-n˝ n j ir-n˝ to:ʃl-i:-ləm Ob prince boy-DLAT rod-DLAT wither-PRS-SG<1SG 'I wither the Ob-prince boy on a rod' (P 1278.22)b. 9k w poŋɣwəl poɒ l-ae wujan j ux-n˝ k w aenə maenət-ow-əs erroneously interpreted as DOs (Section 3).Second, it needs to be justified why Middle Lozva has been excluded from the investigation (Section 4).An additional point of interest is the grammaticalization of the (already burdened) dative-lative as a marker of direct objects.This topic cannot be extensively discussed in this paper, but in Section 6 the traditional idea of a grammaticalization starting from the ablative will be confronted with one which starts from the dative-lative case itself.Section 7, finally, contains the summary and conclusions.The final section of the present introduction (1.4) introduces the data.

Data
The only Mansi variety still spoken is North Mansi (see, for example, Skribnik, Laakso 2022 : 524), whereas the object marking varieties of Mansi are all extinct.Investigations into Mansi DOM must rely on the texts recorded in the nineteenth and early twentieth centuries (Wogulische Volksdichtung 1951-1963;Munkácsi 1892Munkácsi -1896, including texts collected by Antal Reguly).Most of those texts are accessible with glossings and translation in the Ob-Babel Ob-Ugric database (OUDB). 7The database does not include the texts from South Mansi (Tavda), which, apart from their original publication in the volumes by Kannisto and Munkácsi, have been made accessible with (Hungarian) glossings and translation by Norbert Szilágyi.For the present investigation, all West Mansi texts in the Ob-Ugric database -66 from Pelym, seven from North Vagilsk, 31 from Middle and four from Lower Lozva -have been checked for occurrences of dative-lative marked direct objects.The representation of West Mansi in the Ob-Ugric database is not extensive, i.e. it does not contain all West Mansi texts from "Vogul Népköltetési Gyűjtemény" (Munkácsi 1892-1896) and "Wogulische Volksdichtung" (1951)(1952)(1953)(1954)(1955)(1956)(1957)(1958)(1959)(1960)(1961)(1962)(1963).The remaining texts were cursorily checked.Munkácsi (1894) himself had not observed any dative-lative flagging of direct objects in the West Mansi varieties studied by him (Pelymka and Middle and Lower Lozva).In quite a number of instances the actual identification of a DLAT-encoded constituent as a DO is not straightforward.These cases will be examined in Section 3. Lower Lozva does not show any DLAT-encoded DOs, Middle Lozva does, but their status is too uncertain to be included in the present investigation (see Section 4).The investigation finally builds on 25 texts from Pelymka and four from North Vagilsk.This subcorpus contains all texts in which at least one DLAT-DO occurs plus other texts of a mainly narrative character.Excluded are prayers and encantations.The texts in questions come from five different speakers, all of which produce DLAT-DOs (see Table 2).

Speakers, texts, and DLAT-encoded DOs
The text corpus consists of two main types of texts, narratives ("prose") and songs ("poetry/songs").The demarcation of these types is far from straightforward.On the one hand, the narratives are folktales which contain of lot of formulaic and parallelistic structures.The songs, on the other hand, often show narrative structures.In addition, there are smaller forms of folklore (riddles) and descriptional texts (e.g.how to dry meat).None of the texts would document spoken language, be it in form of personal narratives or spontaneous dialogues.More than once a song consists of parallelistic structures, and in some cases, an awkward construction appears paralleling a preceding verse with a more lucid grammar (cf. the discussion of examples (9a) and ( 21c)).Another difficulty may be caused by uncertainties in identifying the referent of an objective conjugation ending (cf. the discussion of (20)).By and large, however, there are no general obstacles to an analysis of the texts in terms of syntax and information structure.

2.1.
A modern cross-dialectal study of Mansi differential object marking does not exist.Early observations come from Munkácsi's dialectal overview (1894), who, for those varieties which possess an accusative, remarks that the marked DO is definite and the unmarked DO indefinite (Munkácsi 1894 : 105 for Middle Lozva, 156-157 for Lower Lozva, 191 for Konda, and 261 for Tavda).DO flagging with the dative-lative case was not observed by him (see Munkácsi 1894 : 235 for Pelymka).He observes an interaction of object marking with possessor marking: object expressions bearing a possessor suffix of the first or second person usually dispense of object marking, whereas those which bear a possessor suffix of the third person may in addition take the accusative suffix.These observations are repeated in later studies (Lavotha 1953;Wickman 1955 : 61-71).Wickman (1955 : 66-69), however, takes a closer look at East Mansi (Konda).He lists clauses with object expressions which do not correspond to the expected pattern, according to which definite objects would be always flagged and indexed, whereas indefinite objects would be neither flagged nor indexed.These earlier descriptions work with isolated sentences, and the fact that indexing has to do with topicality of the object whereas flagging brings in something else was not known at that time.In the next sections, I review later studies.Short summaries of DOM in East and South Mansi can also be found in Bíró, Sipőcz 2018.

DOM in East Mansi
Accusative-based DOM in East Mansi has been investigated mainly by Susanna Virtanen (2013a;2013b;2014;2015 : 39, 41-43).Her research addresses the East Mansi alignment system in general, including also the use of objective conjugation, passive construction and secundative alignment.For object marking in particular, she confirms Munkácsi's (1894 : 191) remark that accusative and possessor marking occur in combination only in the case of the third person singular possessor suffix.Further, she observes that accusativeencoded DOs usually agree, i.e. they are s e c o n d a r y t o p i c s (Nikolaeva 2001;Dalrymple, Nikolaeva 2011 : 53-55), and that the contribution of object marking is twofold: either mere e m p h a s i s (Virtanen 2015 : 42;2013b) or support in contexts of lesser a c c e s s i b i l i t y (Prince 1981;Chafe 1987;Lambrecht 1994 : 100), i.e. topics which are not sufficiently a c t i v e but accessible via the situation (4a) (the door in a room), the context, or inference, or reactivated after some distance between the present clause and their last mention -four sentences in the case of (4b) -are, in addition to being indexed on the verb, also encoded with the accusative (Virtanen 2015 : 42).Virtanen calls this support the s p e c i f y i n g function (Virtanen 2015 : 42;2013b).
(4) a. õõw-mø öät kont-iiløm door-ACC NEG find-SG<1SG 'I cannot find the door' (Virtanen 2015 : 42) b. täw toonøtäätøl nyõõl-mø k°än mänømt-øs-tø 3SG then arrow-ACC out tear-PST-SG<3SG 'Then he tore the arrow out of him' (Virtanen 2015 : 42) Virtanen's idea of object marking as a support device to prepare a DO of insufficient accessibility for agreement is promising and will be taken up below.It is, however, important to note two things.First, with lexically mentioned DOs we find, on the one hand, structures like (4a), which may be topicalizing rather than marking topicality proper (i.e.'The door, I can't find it'), and, on the other hand, proper topical objects as in (4b), whose lexical mention is motivated by distance rather than by first mention.And second, also in East Mansi, DOM is not reserved to agreeing topical DOs.Albeit rarely, accusativemarked DOs do occur with subjective conjugation as, for example, in (5), where, according to Virtanen (2015 : 39), the DO in ( 5) is a contrastive topic.However, the use of subjective conjugation points to a focal DO and I read it rather as a contrastive focal expression, which means that the accusative may also mark accessibility in focal contexts.
(5) näär sons-i sågrøp-mø sons-i what look-3SG axe-ACC look-3SG '[A man and an axe are in one room.]Whatever he (the man) is looking at, it's the axe he's looking at' (Virtanen 2015 : 39)

DOM in South Mansi
South Mansi (Tavda) DOM is also based on the accusative.The variety has been studied by László Honti (1969), Peter Sherwood (1996) and more recently by Norbert Szilágyi (2014).Both Sherwood and Szilágyi observe accusative marking with agreeing and with non-agreeing object expressions, although the latter are rare.These must be focal accessible objects.This is because, as Honti (1969: 120) had already observed, marked objects in Tavda are always definite (határozott), but not all definite objects are marked -d e f i n i t e is here understood as pragmatic definite, i.e. accessible.However, there seems to be a trend to combine object marking with formal expression of accessibility, i.e. actual definite object expressions.According to Szilágyi (2014 : 213-214), South Mansi object marking is frequent with possessor-marked DO expressions, especially with those bearing the possessor suffix of the third person singular.South Mansi thus seems to stand out from the other accusative marking varieties, where the accusative is not used in the case of object nouns bearing a possessor suffix other than that of the third person singular.While definiteness and accessibility are thus meaningful object marking parameters, animacy is considered irrelevant (Szilágy 2014 : 214).
Both Sherwood (1996) and Szilágyi (2014 : 214) bring up the idea of an unsystematic occurrence of object marking in Tavda due to language decay.If this were the case, then the rules of Tavda DOM would not be accessible any longer despite its documentation.It should be noted, however, that we are dealing with a f l u i d type of DOM here which "works solely according to probabilistic rules" (Witzlack-Makarevich, Seržant 2018 : 28), and unpredicted cases of unmarked DOs are not necessarily a sign of language decay.

DOM in Middle and Lower Lozva
These varieties have not been examined recently.Munkácsi (1894 : 156) writes about Lower Lozva: "The indefinite object is unmarked here as it is in Middle Lozva, e.g.uj šª orp äls 'the bear killed an elk' [---].In the same way, the definite object is unmarked if there is a possessor suffix of the first or second person, or of the third person dual or plural".9

DOM in Pelymka and North Vagilsk
Dative-lative based DOM in Pelymka and North Vagilsk has received attention in an article by Angela Marcantonio (1993), which was inspired by work on information structure by Jean Perrot (1994).Marking a DO "has something to do with its 'status as known element' ", but, as she observes, referentiality and k n o w n n e s s are only necessary but not sufficient conditions for the use of object marking and object agreement (Marcantonio 1993 : 34-36).Her expression "known element" may be understood more widely as a c c e s s i b l e or more narrowly as t o p i c a l, i.e. occurring in the presupposed part of a sentence and not in the focus (Lambrecht 1994 : 118;Klumpp, Skribnik 2022: 1019).The notion of focus is not integrated into Marcantonio's account, and the important rule that a focal DO does not agree is not known to her.This does not seem to be a problem because she does not find any DLAT-encoded objects with subjective conjugation (Marcantonio 1993 : 36) (cf. 3.8).What she argues for is a cumulative interpretation of DOA and DOM according to the speaker's "communicative intention": if an element is important in the development of a story, "then the speaker will use all the linguistic devices at his disposal to alert the listener about it" (Marcantonio 1993 : 38-39).In other words, the dative-lative marking is thought to be a device by which the speaker highlights an object expression if (s)he feels that the referent is important (see above Virtanen's (2015 : 42;2013) emphasis function of East Mansi object marking).Such a marking gains importance if the object expression does not occur in its canonical position in front of the verb.So, for (6a), Marcantonio assumes that it is the object marking which allows for the word order VO instead of OV (Marcantonio 1993 : 40).This particular example, as discussed in 3.7, does not display a DO, but a good example for her argument would be (6b).

Summary
Based on the previous studies, the following properties of Mansi DOM can be depicted.First, object marking is not redundant: neither are all agreeing (indexed) objects marked, nor do all marked objects agree.Second, while differential object agreement depends on the discourse topicality of the DO referent (in the understanding of Lambrecht 1994), differential object marking depends on the prominence of the DO referent (in the understanding of Aissen 2003).This is in line with the nature of these morphosyntactic devices: agreement with the verb allows for zero expression of the object referent, which, in turn, is possible only with a salient, topical object referent; marking requires lexical or pronominal expression and applies to prominent objects, possibly topicalized, dislocated, or semi-active (for a cross-linguistic perspective on DOI and DOM, see Iemmolo 2010 : 266).The crucial question for Mansi DOM is why some prominent objects are marked but others are not.Virtanen (2015 : 42) answers with the observation that object marking in East Mansi correlates with insufficient accessibility, i.e. an object expression is accusative-encoded in order to mark the accessibility of its referent after some distance between the last mention and the utterance, or topicalizes the DOs if the referent has not been introduced but is accessible via the situation or context.Marcantonio (1993) is less concrete.In her account, the speaker assigns importance to referents in the discourse, and the more important a referent is the more likely it is that the expression is marked for objecthood.In addition, she connects object marking to dislocation, which is again, very much in line with crosslinguistic observations (Iemmolo 2010).Animacy, on the other hand, plays no role in these accounts.Other negative demarcations are formal definiteness, i.e. object NPs with a demonstrative pronoun and/or a possessor suffix do not require object marking, different from a language like, for example, Komi (see Klumpp 2014 : 419), and the type of the predicate, i.e. notions like partial affectedness or verbal aspect are completely irrelevant for Mansi DOM.It may also be noted that none of the mentioned studies tries to analyse object marking in a d i s c r i m i n a t o r y or d i s a m b i g u a t i n g approach, where object marking is motivated first of all by the need to distinguish a topicworthy object from the subject (cf.Witzlack-Makarevich, Seržant 2018 : 30).Rather, object marking signals properties of the direct object, independently of the properties of the subject.The reason for the dominance of this line of thinking is probably that in Ob-Ugric subjecthood of the most topicworthy referent is guaranteed by the alignment system, and confusion of subject and object is therefore less an issue.
In Section 5, the above mentioned notions will be picked up and checked for their relevance in West Mansi.At this point, a terminological clarification is necessary: in the following I use the terms given and givenness in the tradition of Chafe (1987) instead of accessible and accessibility.

Identifying dative-lative encoded direct objects
3.1.This section is about distinguishing actual DO expressions from other dativelative encoded constituents.Identifying a dative-lative encoded constituent as a direct object is straightforward if this constituent occurs as a second argument in an active clause with a finite verb of high transitivity like, for example, 'kill' and is indexed on the verb by an ending of the objective conjugation (7).
(7) wujaen j s j ˝x-n˝ jal t j i w9:lt-əs-tə bear-DLAT PFV EMPH kill_bear-PST-SG<3SG '[Their uncle went hunting …] He did kill the bear.' (P 1341.54)Often, however, identification is not that straightforward, and for the examination of DLAT-encoded DOs in Section 5 it is necessary to make decisions in favour or against a DO reading.Such cases are discussed in the following section.

Identification by objective conjugation
With verbs like, for example, 'trample', 'hit, or 'lick' one may be inclined to read a DLAT-encoded constituent rather as a goal adverbial ('trample on', 'hit at','lick at') than a DO.If this constituent shows agreement with the verb, such a reading is not appropriate, and the DLAT-encoded argument is a DO (8a-8b).( 8 A number of instances show verbs with very restricted object taking properties, for example 'lie the night' in (9a).If the verb was in subjective conjugation, one may be inclined to read the construction as 'lie down for the night' and identify the DLAT-encoded constituent as a time adverbial rather than a DO.But the objective conjugation ending demands again a DO and there is no other possible DO referent than the night.The construction is possibly motivated by a similar one with a clearer transitive reading, namely 'spend the night' (9b).In the bear song in question, (9b) actually precedes (9a), and the expression 'lie the night' can be understood as parallelistic variation of 'spend the night'.Weak transitivity is also found with DOs which form an etymological figure together with the verb as, for example, 'work the work' or 'run the run'.For the construction in (10), one may consider an alternative reading in which the OC ending on the verb would index a covert DO -some kind of work mentioned earlier -and the DLAT-encoded constituent would be a goal adverbial with a degree meaning: 'I worked (it) (up) to that much work'.However, such a "degree goal" is not a robust category.In Mansi, the notion 'up to, until' is expressed by postpositions like moʃk, and I did not succeed in finding evidence for the dative-lative in such a meaning.

Detached NPs
A DLAT-encoded constituent which appears on the left border of a clause, such as in (14a), may possibly be a detached NP.The detachment is clearer when the rest of the clause is separated by a speech verb as in (14b-14c).
The detached constituent is a topicalized expression, which is coreferential with an OC ending in the following clause.But how should the dativelative encoding of the detached constituent in (14a-14b) be interpreted?Is it object marking, i.e. anticipating the DO function the topicalized argument takes in the following clause?Or is it a marking strategy for topicalized constituents in general ('as for X, …')?Since the same detachment structure occurs without DLAT-encoding (14c), and since I'm not aware of DLAT-encoded left-detached constituents which would appear in the following clause in another syntactic function than DO, I decided for a DO reading in all three cases (14a-14c) (cf.also Marcantonio 1993 : 39).
NEG Q know-PRS-2SG<2SG 'A/the man called Bullfinch, she says, have you not seen him, do you not know him?' (P 1262: 35) Clauses with word order VO sometimes allow for a right detachment reading, such as in (15).In this position, however, identifying a DLAT-encoded constituent as a DO is uncontroversial.

Nonfinite clauses
A DLAT-marked DO can depend on a nonfinite verb (infinitive, participle, gerund).In the case of a promoted recipient DO, as in the first clause of (16a), one may be inclined to read the DLAT-encoded constituent as an IO.However, instrumental encoding of the theme ('vodka') points to a promoted DO in a secundative construction -cf. the same constellation with a finite verb in (16b

Middle Lozva
The Middle and Upper Lozva varieties of West Mansi show accusative encoding of DOs, such as in (19a).As mentioned before (2.4.1), the details of DOM in these poorly documented varieties have not been studied so far.One problem is that accusative encoding also appears with constituents which are not proper DOs, such as in (19b).The example comes from a bear song, and if the detachment structure is due to metric reasons or to actual topicalization, that is beyond my expertise.Apart from accusative vs zero marking, the dative-lative is also found in expressions which are candidates for a DO reading.In the Ob-Babel corpus, the two songs and four invocations from Lower Lozva do not contain any instances, but in the 31 texts from Middle Lozva I found roughly ten instances.A good deal of them can be sorted out by the criteria applied in Section 3, but two instances are strong candidates (cf. already Liimola 1954 : 45-46).First, the noun su:ntəmnə 'opening.1SG.DLAT' in (20) allows for a DO reading if the OC ending refers to it rather than to the hunting path of the preceding clause.In the other reading, the DLAT-encoded constituent would be a translative or goal adverbial, but this interpretation is not attractive, first because Mansi has an actual translative case, and second because the notion of a dativelative encoded goal adverbial has actually been dismissed in 3.2.Thus, the DO reading seems to be the strongest.The other instance of a possible DLAT-encoded DO in Middle Lozva is (21c).The reading depends on the decision about the verb u:-, which allegedly means 'to see, to know' rather than 'to look at' and usually occurs with DOs (see Munkácsi, Kálmán 1986 : 711;Wogulisches Wörterbuch 2013 : 136).The verb is in subjective conjugation, the DO is focal and either new, which is an unusual context for object marking in Mansi (see Section 2), or given.The clause in (21c) is the final verse of a bear song, preceded by two verses which also report what the bear is presented with at the bear feast, where the dead bear, decorated with silver coins and hosted with food and drink, is part of the audience watching performances of dance and song.Both passive clauses in (21a-b) feature the verb ʃunʃt-'to show', and the clause in (21b) is paralleled by the active clause in (21c).Possibly, however, the verb form u:-s-əm may be analysed as u:-wə-s-əm 'see-PASS-PST-1SG', with a contraction of the stem and the passive morpheme, in which case the constructions in (21) would be coherent.12(21) a. n j aur wo:j s j un j  For the purpose of the present study, the Middle Lozva instances are too few and too unclear and therefore will not figure in the remainder of the paper.

Marked vs unmarked given direct objects
5.1.In the following comparison of marked and unmarked direct objects in the Pelymka and North Vagilsk text corpus, all robustly identified DLAT-DOs (see Section 3) plus all other object expressions with a given (accessible) referent figure.Givenness may have different sources as pointed out in Section 5.4.I do not take into account object expressions with new referents or non-referential object expressions, for which object marking is not expected a priori (see Section 2).More than once the folktales contain repetitions of episodes.Some of those repetitions, in which the same clause with exactly the same encodings is repeated for a third or fourth time, were dismissed.Any instance of variation, however, was taken into account.
The investigation deals with lexical DOs only because West Mansi personal pronouns show the same oblique form for direct and indirect objects.This form may have several variants (see the paradigms in Honti 1988 : 155), but the variation does not correlate with different syntactic functions -at least I could not detect such a correlation.Demonstrative and interrogative pronominal DOs are usually unmarked.The only instance of DLAT-encoding of a demonstrative pronoun in DO function I found is ( 22).The function of the pronoun here is resumptive after a topicalizing noun phrase.It is a focal DO expression with subject conjugation.Due to the scarcity of instances, however, pronouns are excluded from the present investigation.
(22) oɒ ʃəx ne: t j it j -n˝ t j i n j ur-eae s-əm?
good woman this-DLAT EMPH want-PST-1SG 'A good woman, is this what I wanted?' (P 1321.49) Altogether I counted 323 DO expressions with a given referent (266 from Pelymka and 57 from North Vagilsk), of which 246 are unmarked (212 in P and 34 in NV) and 77 are .As Table 2 in the Introduction has shown, there is an exceptional high number of 20 DLAT-DO occurrences in text no.1263 "Four Sisters and a Man with his Daughter" by NV speaker V. F. Peršinovna.This text is the reason why the percentage of DLAT-marked DOs in NV (40.5%) is twice as high as in P (20.3%).I have no explanation for why exactly this text is so loaded with DLAT-marked DOs.Still, I resist treating the dialects or even the speakers separately, and the following percentages are presented without reference to dialects or speakers.

Agreement
In the varieties investigated here, it is also most common for DLAT-encoded DOs to occur with objective conjugation, i.e. to agree in number with the verb.This is the case in 71 of 73 instances with a finite verb, whereas among unmarked DOs only 121 of 214 instances with a finite verb agree (see Table 3).The reason for non-agreement of an object is its focality, as was pointed out in 3.8.

Agreement of given lexical DO expressions
The table also shows the occurrences of DOs with non-finite verbs, which make altogether roughly 11% of all instances.Again, this is less common for DLAT-marked DOs (4 out of 77) than for unmarked given DOs (32 out 246), but the difference is not as striking as it is for agreement.

Possessor marking
Possessor suffixes turn a noun phrase into a definite expression and definiteness is a well-known object marking parameter (see, for example, Aissen 2003).Among unmarked given DOs, possessor-marked instances are the majority, almost 62 percent (see Table 4).In turn, absence of a possessor suffix is characteristic for an even larger majority of almost 80 percent of DLAT-marked DOs.The minority of 16 double-marked DO expressions (possessor suffix plus DLAT) includes ten kinship terms, i.e. nouns which almost never occur without a possessor suffix.This distribution is different from the one in South Mansi, where Szilágyi observed frequent object marking of possessor-marked objects (see Section 2.2).It gives reason to consider DLAT-marking a strategy to mark the givenness of a DO referent if there is not already another givenness marker on the expression.This observation is in line with earlier statements on the co-occurrence of possessor an object marking (see Section 2).Table 4 does not indicate which possessor suffixes occur in the data.In the case of the DLAT-marked DOs, these are exclusively third person singular (14 out of 16) (as in other Mansi varieties, see Section 2) and first person singular (2 out of 16), while in the case of the unmarked DOs, these two (82 times 3SG and 44 times 1SG out of 152) are followed by a few occurrences of the other possessor suffixes too.

Type of givenness, distance and syntactic function of antecedent
Givenness, in general, holds for all the referents of object expressions here, with few exceptions as will be shown.I was interested in the sources of givenness, which can be distinguished among the object expressions investigated here.unmarked ( 246  There is a tendency for DLAT-marked DOs to be more frequently given by textual occurrence (42.9%) than unmarked DOs (20.3%), which for the most part are given via a possessor (69.5%) (cf.5.3).The percentage of universally given referents (e.g. the night) is also higher among DLAT-marked DOs (15.6% vs 2.4% for unmarked DOs).Notably, new referents make 10% among the marked DOs, while unmarked DOs add up to only 2.4%.This observation is very much in line with Virtanen's (2015) association of object marking and low activeness at the time of utterance.

Types of givenness among lexical given DOs
For textually given DOs, I also checked the distance and syntactic function of their antecedents (see Table 6).The average number of clauses back for unmarked DOs is 4.8, but for DLAT-marked DOs it is slightly higher, namely 6.3.If that means anything, it is a tendency for marked DOs to allow for a larger distance.However, this parameter can hardly achieve any strong results because we are dealing with lexical DOs and not with pronominal reference or prodrop, where distance may have a much clearer effect.For distances below ten clauses back I counted the previous syntactic function, and here we see a correlation: a transition from the syntactic function of subject to that of direct object occurs with unmarked DOs only in 26.1% of all cases (12 out of 46), whereas with DLAT-marked DOs the percentage is 52.4% (11 out of 21).In turn, a continued direct object function is the most common case for unmarked DOs (23 out of 46, i.e. 50%), whereas this constellation is in second place with DLAT-marked DOs (6 out of 21 i.e. 28.6%).

Type of referent
Animacy has not been identified as a relevant object marking parameter in Mansi (see 2.4).Still, I was interested in whether possibly high animacy (human referents)13 may play a role.For this purpose, I checked object marking for the following types of referents: p r o t a g o n i s t s, k i n s h i p t e r m s, b o d y p a r t s, a n i m a l s a n d h u m a n s w h o a r e n o t p r o t a g o n i s t s, a r t e f a c ts, o b j e c t s o f n a t u r e , s u b s t a n c e s , a c t i o n s, and t e x t.Of these, a n i m a l s a n d h u m a n s as well as the last four groups did not show any remarkable tendencies, and therefore they are summarized in Table 7 as "other".With the remaining groups, however, clear correlations can be observed: DLAT-encoding occurs much less frequently with artefacts, whereas these constitute the biggest group (over 45%) among unmarked DO expressions.The biggest group among DLAT-marked DOs are protagonists, i.e. the group of referents with the highest empathy (48.1% vs 4.5% for unmarked DOs).Also, kinship terms are proportionally more frequent among the DLAT-marked DOs (13% vs 6.1% for unmarked DOs).Another observation is that no body parts figure among the DLAT-marked DOs, whereas this group makes up roughly a fifth (22.3%) of the unmarked DOs.Body parts usually come with a possessor suffix (in 43 of 55 instances) (cf.above 4.3).

Table 7
Types of referents among lexical given DOs Having a closer look at the six DLAT-encoded artefact-DOs reveals that in three instances the DO referent is a fish weir which plays a central role in the texts in questions.In two instances, it occurs with subjective conjugation (1289.22-23, see (18), and 1289.47-49) and once with a nonfinite verb (1264.014).The remaining three instances of DLAT-encoded artefact-DOs come with OC: unmarked (55)   two times it is a knapsack, which is important in the story because some protagonists are hidden in it (1263.088and 1263.098); and in one instance it is a crowbar (1263.054),which is used four times to pierce a victim (1263.015,035, 054, and 078), but only once follows the explicit comment that the victim is killed, and in this instance the mention of the crowbar comes with DLAT-encoding (24a).Similar observations can be made for the category of substances (earth, tears, vodka), where we also find object marking if the referent is somehow important, e.g. the earth in (24b), which occurs as the agent in a passive clause and then as a DO.Animacy thus proves to be a relevant parameter in that DO expressions which range low on the animacy scale need to play an important role in the plot in order to be object-marked.

Type of DO
Promoted DOs are semantically recipients or locative expressions of a three argument predication.The promotion includes their obligatory agreement with the verb and, if there is a patient argument, this is encoded with the instrumental case.DLAT-encoding of a promoted recipient, as in (25), looks as if, in the promotion process, the DLAT-marking has been "forgotten" to be removed from the recipient argument.From a diachronic point of view, one may consider this construction to be a source of dative-lative encoding of direct objects (see Section 6).In the text corpus investigated here, promoted DOs are rare (only 4%), but the majority of these are object-marked, 11 vs 2 unmarked ones (see Table 8).A small number of marked promoted DOs is basically expected because promotion is due to high topicality, which leads to zero anaphora rather than to lexical expression.

Word order
The unmarked word order in Mansi is OV (see Rombandeeva 1984 : 85-60;Bíró, Sipőcz 2018 : 14;Virtanen 2021 : 209, 229;Forsberg 2022 : 575), and Marcantonio (1993 : 39-40), with good reason, wondered if object marking may correlate with a change from OV to VO (see 2.3).Indeed, the percentage of VO clauses with DLAT-marked DOs is slightly higher than with unmarked DOs.Table 9 is simplified and does not show all registered word order variations (oxv, ovx, xov etc.), but I added the numbers for instances with an overt subject expression.Such co-occurrence may have an effect on object marking if there is a need to disambiguate the DO from the S (see also 5.7).However, for DLATmarked DOs the percentage is only slightly higher, and this hardly gives reason to consider the overt presence of a subject expression to be a relevant object marking parameter in West Mansi.However, I also checked whether the clause a given DO occurs in appears after a transition from one text level to another, and found that object marking occurs a bit more often after such a transition (36.4% vs 24.4% for unmarked DOs), which I consider a significant result.

Table 10
Lexical given DOs and text level

Discussion
The frequencies as presented in the preceding subsections confirm a couple of earlier observations and assumptions about DOM in Mansi.First, object marking applies to lexical DO expressions with a given referent, and second, marked objects almost always agree, while only about one half of the unmarked objects do.DLAT-encoded DOs are also clearly less frequently possessor marked.Some of the observations for West Mansi bring new aspects into the discussion.First, object marking clearly occurs more often with referents of very high animacy, i.e. protagonists, and less often with artefacts.Second, for marked DOs textual givenness is the most frequent source of givenness, but for unmarked ones it is givenness via a possessor.Third, the ratio of promoted DOs is higher among DLAT-marked DOs.And fourth, marked DOs occur more easily (frequently) in "challenging" contexts.These are: distant last mention, transitions in syntactic function and text level.
The above investigated parameters also confirm Virtanen's (2015) observation for East Mansi that object marking occurs often in contexts of semiactiveness, and in addition to lexical mention, object marking supports the agreement of these object expressions.Typical activation contexts are distant last mention (26a-26b), transition from one text level to another (27a-27b), frame-based givenness (28a-28b), or progressive anaphora (i.e.different expressions for the same referent) (29a-29b) (see e.g. Grüning, Kibrik 2002).In all these cases, however, we find not only marked but also unmarked DO expressions and additional arguments are needed in order to explain why, in addition to lexical expression, object marking is needed or not.For ( 26) and ( 27) one may argue in terms of high animacy (empathy) that the marked DOs in the a-clauses (the girl, the daughter) are protagonists while those in the b-clauses (the pin-throwers, the legs) are not.For (28) the reason may be that in the riddle question in the a-clause the DO (the black bottomed wolf as a metaphoric enigmatization of a cauldron above the fire) is less accessible than the DO in the b-clause (the door of the hut).For (29), however, I cannot think of a reason why the coreferential expression 'girl' (after 'sister') would require object marking but 'boy' after 'son' would not.
'The two other wives became envious that she had borne a son: she will be loved, they won't, they haven't borne a son.The woman who had given birth fell asleep, the other two stole the boy and threw him into the sea.' (P 1264.53-54)Thus, there are parameters which favour or disfavour West Mansi object marking, but there seems to be no absolute rule according to which a DO expression with a certain property in a certain constellation would need obligatory object marking.If there were a larger corpus, it would be interesting to calculate parameter results in relation to each other and look for statistical significance.With the small corpus at hand, however, I doubt that this will achieve results.After all, it seems that we do not get far beyond Marcantonio's (1993) idea that speakers mark an expression when they think the referent is important for the development of the story.We are left with free choice by the speaker.

Origin
Since Liimola (1954 : 43-45), the explanation for the DO-marking function of the Pelymka and North Vagilsk dative-lative case starts from the ablative.After the merger of the ablative (-nəl) and the dative-lative (-nə), separative adverbials in constructions with the verb 'to fear' ('be afraid of'), with affected body parts (e.g.'take at [lit.from] the hand', see 3.4), and with partitive meaning (e.g.'ate from the food') were reinterpreted as direct objects, then dative-lative marking was generalized and replaced accusative marking.In support of this assumption, Liimola (1963 : 45) refers to the history of the Finnish partitive case, which goes back to the Proto-Uralic ablative.Wickman (1954 : 64) finds "a still more striking parallel" in Saami, where the former partitive has ended up as a marker of definite plural objects.Later authors refer to this explanation until the present day (Riese 1992;Honti 1988 : 153;Marcantonio 1993 : 28), whereby Liimola's intelligent idea of reinterpretation of separative adverbials was reduced to the "partitive function of the ablative" only.Partial objects, however, do not play a central role in any of the Ugric languages.In the following, I would like to challenge the traditional explanation with one which starts from the dativelative, independently of its separative meaning due to the merger with the ablative.
Languages in which a dative case marks DOs are not rare (see, for example, Iemmolo 2013), and "dative-accusative syncretism" (Naess 2008 : 578) is linked to a greater degree of independence and individuation of marked direct objects, making them similar to typical dative marked participants.For West Mansi, where the case in question is a dative-lative, whose most frequent function is the marking of usually inanimate goals, this explanation needs additional support.Iemmolo (2011 : 15-28;2013) links the dative and lative (allative) to topic marking constructions (as for X, …), though this is not necessarily the starting point for grammaticalization into a direct object marker.Topicalization was mentioned with left-detached constituents in 3.6.I'm not aware of a Mansi topicalizing construction which would regularly apply the dativelative, and therefore I'm reluctant to identify cases like (14a-14b) in 3.6 as the source-construction for the grammaticalization of the dative-lative into a marker of direct objects.
A second source was considered in 5.6: a topical recipient or addressee in being a promoted direct object with number agreement with the verb preserves its originally semantically motivated DLAT-encoding.Promoted DOs with lexical expression, however, were found to be rather rare, and it is questionable if this is the (only) source of grammaticalization.
A third idea relevant for the grammaticalization contrasts the direct object marking function of the dative-lative case to its older, cross-dialectal agent marking function in passive clauses and with participles.Agents are circumstantial constituents, which are obligatorily marked, and thus differ from direct objects, which are core participants and marked differentially.But both agents and marked objects are non-subject constituents with a high likelihood of ranging high on the animacy scale, and they occur in complementary distribution: agents in passive clauses and direct objects in active clauses.For both, the most likely confusion, namely confusion with the goal function, which is the most frequent funcion of DLAT-encoded constituents (see 1.2), is reduced by their non-prototypical goal properties.For example in (30), the two DLAT-encoded constituents, the floor and the knife are both inanimate, but the goal-constituent is prototypical, whereas the knife is a (magic) self-acting tool and therefore prone to occur in a syntactic function different from a goal.In the first case, the objective conjugation ending in the following verb refers to the covert DO elk, whereas in the second case the ending refers to the DLAT-encoded DO.
(30) o:s sus kaŋəlpeae t j ət-tə oɒ ləmt-əs k w aelkaen-n˝ rast-əs-tə again elk armpit-LOC carry-PST floor-DLAT throw-PST-SG<3SG jeaept-n˝ rast-əs-tə knife-DLAT throw-PST-SG<3SG 'Again, he was carrying an elk under his arm, he threw it onto the floor, he threw the knife' (P 1262.47)There are thus three possible factors to consider for the dative-lative encoding of direct objects in West Mansi: topicalization, indirect object promotion, and marking transfer from the passive construction.Relevant for the grammaticalization of the dative-lative as a marker of direct object is not only the transparent adverbial meaning the dative-lative case has with goals and recipients and addressees, but also the non-subject marking function the dative-lative case has with agents in passive clauses.The link between the two is supported by the fact that a DLAT-encoded DO and a DLAT-encoded agent cannot occur in the same clause.A further supporting factor is that the distinction of DO and IO via case marking is irrelevant in Mansi because in ditransitive constructions (S -DO -IO) the DO position is reserved to either of them depending on their topicality status.A more detailed investigation of this question is beyond the scope of this paper.

Summary and conclusions
The aim of this article was to clarify the extent of dative-lative based DOM in the West Mansi varieties Pelymka and North Vagilsk.It builds on previous studies on the Ob-Ugric alignment system and the role of object marking in relation to object agreement in varieties of Mansi.The basic split in object marking is between personal pronouns, which have a distinct object form in all instances, and other nouns, which may show asymmetrical differential object marking (a typologically frequent split; see for example Witzlack-Makarevich, Seržant 2018 : 7-8).The split among the other nouns is between new objects, which are never marked, and given objects, which may be marked.Unlike South and East Mansi, where the object-marking case is the accusative, in Pelymka and North Vagilsk there is no accusative case, but the object marking case is the multi-functional dative-lative case.This peculiarity of Pelymka and North Vagilsk creates more than once the difficulty that a DLAT-encoded constituent might erroneously be read as a DO.In Section 3, criteria were established to tackle this problem.According to these criteria, the number of possible dative-lative DOs in the West Mansi accusative variety of Middle Lozva is close to zero, and this variety has therefore been excluded from the present investigation (see Section 4).
The ratio of unmarked vs marked DOs with a given referent which was detected in the Ob-Babel Pelymka and North Vagilsk text corpus is approximately 3:1 (246 : 77).Among DLAT-encoded constituents, direct objects (77) are the third most frequent group after goal expressions (302) and agents in passive clauses and with participles (134) and before indirect objects (26) (see 1.2).Based on previous observations about Mansi DOM (see Section 2), a number of parameters were selected and checked for their effect on object marking.Object agreement (indexing) occurs more frequently with marked objects (71 out 73, i.e. roughly 97%) than with unmarked objects (121 out 214, i.e. roughly 57%).However, the basic principle that a topical DO agrees whereas a focal DO does not holds also in West Mansi, and the ratio of 121 unmarked and 71 marked DOs among agreeing objects shows that that object agreement is not an effective parameter in predicting object marking (see 5.2).Possessor marking seems to have a negative effect on object marking.While roughly 62% of the unmarked DOs bear a possessor suffix, only roughly 21% of marked DOs do so.This distribution differs from South Mansi, where object marking was frequently observed on possessor marked nouns, and it gives reason to consider DLAT-marking a strategy to mark the givenness of a DO referent if givenness is not already marked by a possessor suffix (see 5.3).Association with a possessor is also the most frequent source of givenness with unmarked DOs (roughly 70%), whereas for marked DOs it is textual occurrence (roughly 43%).Marked objects are also more often universally given (roughly 15.5%) than unmarked objects (roughly 2.5%) (see 5.4).Counting clauses between a DO expression and its anaphoric antecedent did not achieve significant numbers, but for distances between one and nine clauses back I counted possible transitions between syntactic functions from the antecedent to the DO.And here object marking occurs more frequently with DOs which were a subject in their preceding mention (roughly 52.5%) than with continued DOs (roughly 28.5%), whereas for unmarked DOs the ratio is the opposite (50% continued DOs and roughly 26% switched from subject) (see 5.4).Animacy has never been claimed to be a relevant object marking parameter in Mansi, but correlations for high and low animacy can still be observed in West Mansi: the biggest group among marked DOs are protagonists (roughly 48%) and the smallest artefacts (roughly 8%).For unmarked DOs the ratio is again the opposite (4.5% protagonists vs 45.5% artefacts) (see 5.5).Another clear correlation shows in object marking with promoted DOs.Promoted DOs tend to occur with zero anaphora rather than with lexical expression, but among the lexical promoted DOs the majority is marked (11 out of 13) (see 5.6).
The last three parameters fall under the domain of "challenging contexts".Where DO and S coincide in person and number, confusion may arise as to which referent is the S and which the DO.However, disambiguation does not seem to be an effective parameter (see 5.7).Non-canonical occurrence of a DO after the verb (VO) has a slightly higher rate of marked objects (roughly 10.5%) over unmarked objects (4.5%) (see 5.8).The same holds for transitions between different text levels, where roughly 36.5% of the marked DOs occur after such a transition but only roughly 24.5% of the unmarked DOs (see 5.9).
In the discussion of the results, I demonstrated how the different parameters may be effective or not for objects in comparable environments, namely contexts of distant last mention, text level transition, frame-based givenness, or progressive anaphora.These contexts, in general, favour object marking, but for all cases it is possible to find unmarked object expressions too, and also the application of additional parameters like animacy does not lead to a robust object marking rule (see 5.10).The conclusion is that DOM in West Mansi is of a fluid type which "works solely according to probabilistic rules" (Witzlack-Makarevich, Seržant 2018 : 28).Or, put differently, DOM is a device which allows the speaker "to alert the listener" about an element which is considered important in the development of a narrative (Marcantonio 1993 : 38).While the results for West Mansi are much in line with observations about object marking in East Mansi by Virtanen (2015), it goes without saying that this conclusion does not need to be a final one.

Table 1 .
Most frequent are goal expressions, followed by agents of passive clauses and participles, 6 direct objects and finally indirect objects (addressees and recipients): the 77 instances of differential DO-marking clearly outweigh the 26 instances of canonical IO-marking by approximately two thirds.

Table 1 Frequencies of syntactic functions of DLAT-marked constituents in the Ob-Babel Pelymka and North Vagilsk text corpus 1.3. Aim and structure of the paper
The present study aims at understanding the extent of dative-lative based DOM in the West Mansi varieties Pelymka and North Vagilsk, as well as the relevant parameters involved.For this purpose, I examine object marking according to the parameters o b j e c t a g r e e m e n t (i n d e x i n g), p o s s e s s o r m a r k i n g, t y p e o f g i v e n n e s s, d i s t a n c e a n d s y n t a c t i c f u n c t i o n o f a n t e c e d e n t, a n i m a c y, o b j e c t p r o m o t i o n, d i s a m b i g u a t i o n, w o r d o r d e r, and t e x t l e v e l.The results of this investigation are presented and discussed in Section 5.The actual investigation is preceded by a summary of earlier studies on DOM in different Mansi varieties (Section 2) and two clarifying sections.First, actual direct objects need to be distinguished from other dative-lative encoded constituents, which may be (and have been) OnDative-Lative Encoded Direct Objects...
) a. an j s j əx o:s j t-əs b. ke:lp s j e:s j e:m˝l puj-p˝ s j e:s j -n˝ n j alləmt-i-tə red wolf black bottom-ADJ wolf-DLAT lick-PRS-SG<3SG 'A red wolf licks a black-bottomed wolf'(NV 1269.11-12) Across Mansi dialects, addressees of 'ask' are encoded either as DO(Munkácsi,  Kálmán 1986 : 216, Wogulisches Wörterbuch 2013 : 397-398)or with the ablative ('ask from X'), where the addressee figures as a source of information.In West Mansi I have not come across an instance in which the addressee of a question would be unambiguously encoded as a direct object.Are they to be read as DOs as the original translation implies ('she tied his hands very thick, his feet very thick') or as goal expressions ('she tied it [the wire] to his hands very thick, to his feet very thick') or as separative adverbials ('she tied it [him] from his hands very thick, from his feet very thick')?In those Mansi varieties which have an ablative, this case is used in encoding affected body parts as, for example, in (12b) from East Mansi.In view of the merger of dative-lative and ablative in West Mansi, I decided for a separative reading of DLAT-encoded constituents as in (12a), i.e. they are not DOs.Now it's time to tie his feet and hands with copper wire.She tied and tied him with the copper wire, tied him very thick at his hands, very thick at his feet'(P 1258.63-64) 10There are instances of OC, such as kitəl-i:-tə 'ask-PRS-SG<3SG'(P 1260: 78), for which the Ob-Babel corpus provides the translation 'she asked her', but they can also mean 'asks it', i.e. the OC ending refers to the content rather than to the addressee of the question.With subjective conjugation as in (11a-11b), we are clearly not dealing with DOs but with source adverbials.(11)a.m9:kəmkitəl-9ɣ-tne:¸-˝¸-n˝:maenərkomeae l-əl people ask-PRS-3PL woman-DU-DLAT what_kind_of coffin-INS woɒ r-x law-s-əɣ make-INF order-PST-3DU 'The people ask from the two wives, what kind of coffin did they order to be made' (NV 1271.05) b. aŋkəʃk kiteae l-i an j s j ˝x-n˝: [---] old_woman ask-PRS old_man-DLAT 'The old woman asks from the old man [---]' (P 1262: 76)3.4.Affected body partsThe presence of objective conjugation is not always a sufficient criterion for identifying a DLAT-encoded constituent as a DO if it is not obvious which overt or covert argument actually agrees.11Inthethirdclause of (12a), the DLAT-encoded constituents 'hands' and 'feet' follow the verb 'tie', which bears an OC ending (referring to a singular object though).b. om tae:wə wə-s-ləm kit l◊jl-ø ae-n˝l n j ərəmt-əs-ləm 1SG 3SG.ACC take-PST-SG<1SG two foot-DU-ABL grab-PST-SG<1SG 'I took it, grabbed it by its two legs'(EM 1557.16) Now it's time for his mother to give her son vodka to drink'(P 1258.70)Objectiveconjugation was introduced in 2.1 as a criterion to identify a DLATmarked constituent as a DO and most DLAT-DOs, in fact, occur with OC.But there are few instances with subjective conjugation.Recall that OC is a device to mark the topicality of a DO expression next to a topical subject.Topicality entails givenness of a referent, be it textually or frame-based (see more in Section 5.5).Givenness, on the other hand, is a property not only of topics but possibly also of focal DOs, and a focal DO is not indexed (see, for example, Nikolaeva 2001 : 17-19).Such a case is (18), in which a weir maker asks why of all weirs somebody would touch (i.e.steal fish from) his miserable weir.The DO is contrastive focal (m y weir in opposition to all other weirs) and therefore cannot be indexed by an OC ending despite the givenness of its referent.

Table 4 Possessor marking among given lexical DOs These
are: g i v e n v i a a p o s s e s s o r, t e x t u a l l y, u n i v e rs a l l y, f r a m e -b a s e d, s i t u a t i o n -b a s e d (in direct speech contexts) g i v e n, and n o n -g i v e n (n e w)(cf.e.g.Hawkins 1978; Fraurud 1990).The latter category may appear surprising because DOs with new referents are not part of the investigation.However, some instances needed to be included in which a non-given DO agrees with the verb, i.e. the speaker treats them as given without proper introduction.Such a treatment is usually cataphoric, i.e. the referent plays some role in the further development of the text, such as in(23).In other instances, a textually given referent was classified as new because it occurred in direct speech where it was new to the addressee.

Table 8 Primary and promoted lexical given DOs
Disambiguation has not been mentioned as a relevant parameter of Mansi DOM so far.The parameter possibly deserves a more elaborated treatment; for the present investigation I only checked if coincidence of subject and object in person and number may have an effect on object marking.The hypothesis is that in such cases we find object marking significantly more frequently, but this is not the case: of the 77 DLAT-encoded DOs, 33 (42.9%) share person and number with the subject (all third person singular) and 40 (51.9%)differ in person number (e.g.S: 3SG ≠ DO: 3PL, or S:1SG ≠ DO: 3SG).In the remaining four instances (5.2%), the DO occurs with a non-finite verb where no confusion with the subject is possible.That means that in the majority of cases object marking occurs in a context in which the verb and the subject together leave no room for confusing the object with the subject, from which I conclude that discrimination is not the function of DOM in West Mansi.The rates for unmarked objects are not much different and confirm the irrelevancy of the parameter in question: of 246 unmarked DOs, 78 (31.7%) share either 3SG or 3DU, 26 (10.6%) occur with non-finites, and 142 (57.7%) occur together with a subject differing in person and/or number.

Table 9
The latter category does not contain any given DOs.As Table10shows, the three levels are more or less equal in terms of object marking.