Nonstandard Use of the “Reflexive” Affix -sja in Russian Speech of Bilingual Speakers of Northern Siberia and the Russian Far East

One of the features of the oral Russian speech of bilingual speakers of the indigenous languages of Russia is the omission/the overuse of the “reflexive” affix -sja (a “middle voice” marker with a wide range of uses including reflexive, reciprocal, anticausative, passive, and some others). We discuss the data on the nonstandard use of -sja in the Russian speech of bilingual speakers of two language groups that differ both from Russian and from each other in this grammatical domain: Samoyedic (Forest Enets, Nganasan, and Nenets) and Tungusic (Nanai and Ulch). The data come from the corpus of contact-influenced Russian speech, which is being created by our team. We show that the mismatches in standard and nonstandard usage cannot be explained by direct structural copying from the donor language (indigenous) to the recipient one (the local variety of Russian). Nor is there a consistent system which differs from standard Russian since there are many more usages that follow the rules of standard Russian. The influence of the indigenous languages explains some overuses and omissions; the others can be explained by other factors, e.g., difficulties in the acquisition of verb pairs with non-transparent semantic or syntactic relations.


Introduction
One of the features of the oral Russian speech of bilingual speakers of the indigenous languages of Russia is the nonstandard use of the "reflexive" affix -s j a, which can occur as an overuse (1) or as an omission (2). We do not observe notable differences betwe this might be partly explained by the extremely Samoyedic sample.
We divided all nonstandard uses of -s j a into tw -s j a) and overuses (the unexpected presence of -s Samoyedic one, omissions were more frequent expectations on the influence of the indigenous sy expected, unless the correlate of -s j a in the source la say-PRS. 3PL wave-SEM-PST.M.SG 'They say then he swam across the river and waved his hand (to show that he should go well)'. (L1 Nganasan) In (8), a non-standard verb obitat j -s j a is used instead of the Russian verb obitatʲ 'dwell'.

8.
by-va-jut=to, vot, oni tam I be-IPFV-PRS. 3PL=PTCL here 3PL there and obita-jut-sʲa, eto, kormʲ-at-sʲa dwell-PRS.3PL-SJA this.N feed-PRS.3PL-SJA 'There are… so, they dwell there, feed themselves'. (L1 Enets) Enets stative verbs like 'dwell' normally do not bear reflexive suffixes, so this cannot be a ca There is a synonymous Russian verb vodit j -s j a 'be found, live' which has -s j a. However, one c propose a different explanation: the verb kormjats j a in the right context triggers the -s j a suffix o verb obitatʲ.
The verb torgovalis j in (9) is used in the object impersonal meaning ('to sell' > 'to sell diff things').
9. nu ran j she zhe kitajtsy zdes j torgova-l-i-s j PTCL earlier PTCL chinese.PL here sale-PST-PL-REFL 'Well, earlier, Chinese sold different things here'. (Nanai corpus) This particular derived form is absent in standard Russian, despite the object imper meaning is one of the productive meanings of -s j a. So, in Tungusic Russian, we are dealing wit overgeneralization of this meaning. It is not a calque from the indigenous language (Nanai), sin Nanai there is no affix with this meaning at all.

Omissions of -sʲa
It is more difficult to explain an unexpected absence of -sʲa than its overuse. On the one h some omissions can be explained by the absence of s j a-type markers in the corresponding indige language. For example, the form rodila instead of rodilasʲ 'to be born' in Nanai Russian is suppo by the Nanai verb bal ǯ e-'to be born' which is not connected to the verb 'to give birth', unlik Russian correlate.
Thus, both Samoyedic and Tungusic markers overlap with -s j a in the mediopassive semantic domain, but not in the reflexive one. Within this domain, the anticausative suffix -p in Tungusic is similar to the Russian -s j a in terms of productivity, while the Samoyedic reflexive conjugation is much more restricted and it overlaps with -s j a only for a closed set of verbs.

Materials and Methods
As the data source, we used the corpus of contact-influenced Russian speech of Northern Siberia and the Russian Far East, which is being created by our team. This is a small spoken corpus of narratives in Russian recorded from speakers of indigenous languages of the area. The texts are transcribed in standard Russian orthography in ELAN 2 and manually annotated of grammatical and lexical contact-induced features (one of them is the nonstandard use or omission of the reflexive affix). In the study, we used the transcribed and annotated part of the Tungusic and Samoyedic subcorpora, which contained approximately 17 h (29,283 clauses). The whole collection of the records from the speakers of the Tungusic and Samoyedic languages contained approximately 96 h; see Table 1 for the details. The nonstandard use of -s j a is not very frequent in our data (the most frequent grammatical peculiarities are the omission of prepositions and gender disagreement). In our annotated corpus, we found 71 cases of nonstandard uses of -s j a in total:46 uses in Tungusic subcorpus and 25 uses in Samoyedic subcorpus. Only 6% out of all uses of -s j a are nonstandard in Tungusic subcorpus (733 uses). The data, available at the moment, are not enough for consistent quantitative analysis. So, in this paper, we present a preliminary qualitative study, in which we analyzed possible factors that could have influenced the nonstandard uses of -s j a attested in the corpus.

Results
Some of the attested nonstandard uses of -s j a, both omissions and overuses, can be indeed explained by the influence of indigenous languages. However, other cases seem to contradict this hypothesis. We show that some other factors connected to acquisition difficulties can affect the use of -s j a as well. First, we discuss overuses, which are easier to explain (Section 3.1), and then omissions 2 ELAN (https://tla.mpi.nl/tools/tla-tools/elan/) is one of the annotation tools developed at the Max Planck Institute for Psycholinguistics, see (Sloetjes and Wittenburg 2008).
Languages 2019, 4, 39 4 of 9 (Section 3.2). In Section 3.3, we give some quantitative data on the rate of omissions vs. overuses and the distribution across different semantic types of -s j a uses.

Overuses of -s j a
Overuses of -s j a can be classified into three groups: (a) structural borrowing from the indigenous language (PAT-borrowing in terms of (Sakel 2007)); (b) incomplete acquisition of standard Russian: the existence of a particular Russian verb similar to that in question or the overgeneralization of a particular semantic type of s j a-uses.

Structural Borrowing
Examples (3)-(4) illustrate structural borrowing, which leads to the overuse of -s j a.
(3) is a calque from the Nanai impersonal construction (4). The affix -s j a can have the meaning presented in (3)-(4); however, the argument encoding differs from that of the Russian s j a-verbs and repeats that of the Nanai impersonal forms: the direct object takuju 'such a thing.f' does not move to the subject position and takes the accusative case, in the same way asčolombani 'soup.ACC.3SG' in (4). So, -s j a appears in (3) as an equivalent of the Nanai impersonal suffix -wu. Cf. the Nanai impersonal construction in (4):

oaktačolom-ba-ni
Xon' puju-u-r' wormwood soup-ACC-3SG how cook-IMPS-PRS 'How does one cook wormwood soup'? (Nanai corpus) Example (5) is more complicated. The Russian verb sn j at j 'take off' does not take the anticausative -s j a in Russian monolinguals. The overuse of sn j at j -s j a in (5) corresponds to the Nanai ačo-p 'come taken off' [take.off-DECAUS]. However, sn j at j -s j a inherits not only the Nanai morphological pattern but also the lexical one. In (5), it has the meaning 'to come untied' and not 'to come taken off'. This is explained by the fact that its correlate ačo--ačo-p is polysemous:along with the meaning 'to take off-to come/be taken off', it has another meaning 'to untie-to come/be untied' 3 .

Incomplete Acquisition of Russian
One of the factors that can affect the use of -s j a, besides the systems of the indigenous language, is the existence of Russian verbs that have a synonymous meaning but behave differently with respect to the -s j a derivation.
In (1) repeated below as (6), a possible contamination with the Russian reflexive verb t j erjat js j a 'get lost' might have played a part in using propadaj-s j a instead of propadaj. The lexical calquing might be even more important here, because the verb otvazat j -s j a 'to come untied' [untie-REFL] does exist in monolinguals' Russian, in contrast to *sn j at j -s j a. In a similar manner, in (7), there could have been contamination with the Russian verb perepravi-l-s j a (cross-PST.M.SG-SJA). 7. potom, govor j -at, pereply-l-s j a i ruk-oj, later say-PRS.3PL cross-PST.M. SG-SJA and hand-INS govorj-at, mah-nu-l say-PRS.3PL wave-SEM-PST.M.SG 'They say then he swam across the river and waved his hand (to show that he should go as well)'. (L1 Nganasan) In (8), a non-standard verb obitat j -s j a is used instead of the Russian verb obitat j 'dwell'. Enets stative verbs like 'dwell' normally do not bear reflexive suffixes, so this cannot be a calque. There is a synonymous Russian verb vodit j -s j a 'be found, live' which has -s j a. However, one could propose a different explanation: the verb kormjats j a in the right context triggers the -s j a suffix on the verb obitat j .
The verb torgovalis j in (9) is used in the object impersonal meaning ('to sell' > 'to sell different things'). This particular derived form is absent in standard Russian, despite the object impersonal meaning is one of the productive meanings of -s j a. So, in Tungusic Russian, we are dealing with the overgeneralization of this meaning. It is not a calque from the indigenous language (Nanai), since in Nanai there is no affix with this meaning at all.

Omissions of -s j a
It is more difficult to explain an unexpected absence of -s j a than its overuse. On the one hand, some omissions can be explained by the absence of s j a-type markers in the corresponding indigenous language. For example, the form rodila instead of rodilas j 'to be born' in Nanai Russian is supported by the Nanai verb bal  Enets stative verbs like 'dwell' normally do not bear reflexive suffixes, so this cannot be a calque. There is a synonymous Russian verb vodit j -s j a 'be found, live' which has -s j a. However, one could propose a different explanation: the verb kormjats j a in the right context triggers the -s j a suffix on the verb obitatʲ.
The verb torgovalis j in (9) is used in the object impersonal meaning ('to sell' > 'to sell different things').
9. nu ran j she zhe kitajtsy zdes j torgova-l-i-s j PTCL earlier PTCL chinese.PL here sale-PST-PL-REFL 'Well, earlier, Chinese sold different things here'. (Nanai corpus) This particular derived form is absent in standard Russian, despite the object impersonal meaning is one of the productive meanings of -s j a. So, in Tungusic Russian, we are dealing with the overgeneralization of this meaning. It is not a calque from the indigenous language (Nanai), since in Nanai there is no affix with this meaning at all.

Omissions of -sʲa
It is more difficult to explain an unexpected absence of -sʲa than its overuse. On the one hand, some omissions can be explained by the absence of s j a-type markers in the corresponding indigenous language. For example, the form rodila instead of rodilasʲ 'to be born' in Nanai Russian is supported by the Nanai verb bal ǯ e-'to be born' which is not connected to the verb 'to give birth', unlike its Russian correlate.

mī
Bal ǯ e-xam-bi Muxu-du 1SG be.born-PST-1SG Muhu-DAT 'I was born in the village of Muxu.' (Nanai, field records) Example (12) can have a similar explanation; the absence of the s j a-type marker (the reflexive conjugation). See (12) with no -s j a and no reflexive conjugation in the corresponding Nganasan verb from the parallel version of the Russian text (13) Enets stative verbs like 'dwell' normally do not bear reflexive suffixes, so this cannot be a calque. here is a synonymous Russian verb vodit j -s j a 'be found, live' which has -s j a. However, one could ropose a different explanation: the verb kormjats j a in the right context triggers the -s j a suffix on the erb obitatʲ.
The verb torgovalis j in (9) is used in the object impersonal meaning ('to sell' > 'to sell different ings'). nu ran j she zhe kitajtsy zdes j torgova-l-i-s j PTCL earlier PTCL chinese.PL here sale-PST-PL-REFL 'Well, earlier, Chinese sold different things here'. (Nanai corpus) This particular derived form is absent in standard Russian, despite the object impersonal eaning is one of the productive meanings of -s j a. So, in Tungusic Russian, we are dealing with the ergeneralization of this meaning. It is not a calque from the indigenous language (Nanai), since in anai there is no affix with this meaning at all.

Omissions of -sʲa
It is more difficult to explain an unexpected absence of -sʲa than its overuse. On the one hand, me omissions can be explained by the absence of s j a-type markers in the corresponding indigenous nguage. For example, the form rodila instead of rodilasʲ 'to be born' in Nanai Russian is supported the Nanai verb balǯe-'to be born' which is not connected to the verb 'to give birth', unlike its ussian correlate.
. ja zhe derevne rodi-l-a 1SG PTCL village.LOC be.born-PST-F 'Actually, I was born in the village.' (L1 Nanai) . mī Bal ǯ e-xam-bi Muxu-du 1SG be.born-PST-1SG Muhu-DAT 'I was born in the village of Muxu.' (Nanai, field records) Example (12) can have a similar explanation; the absence of the s j a-type marker (the reflexive njugation). See (12) with no -s j a and no reflexive conjugation in the corresponding Nganasan verb om the parallel version of the Russian text (13). be.born-PST-1SG Muhu-DAT 'I was born in the village of Muxu.' (Nanai, field records) Example (12) can have a similar explanation; the absence of the s j a-type marker (the reflexive conjugation). See (12) with no -s j a and no reflexive conjugation in the corresponding Nganasan verb from the parallel version of the Russian text (13). differs from standard Russian. Moreover, sometimes we witness a variation: -sʲa can be omitted and used correctly within one paragraph or even within one sentence, as in (17) We do not observe notable differences between the Samoyedic and Tungusic data. However, this might be partly explained by the extremely small number of nonstandard -s j a uses in the Samoyedic sample.
We divided all nonstandard uses of -s j a into two groups: omissions (the unexpected absence of -s j a) and overuses (the unexpected presence of -s j a). Both in the Tungusic text sample and in the Samoyedic one, omissions were more frequent than overuses. This generally agrees with our expectations on the influence of the indigenous system. The prevalence of overuses indeed is not expected, unless the correlate of -s j a in the source language was much more productive. This was not the case either in Tungusic or in Samoyedic.
However, if we exclude deponent verbs, for which omission is logically the only option, and irregular s j a-derivates, for which overuses are not attested either, overuses, in contrast, become even more frequent than omissions.
Not all nonstandard uses of -s j a are caused by structural borrowing. There are even more cases that can be interpreted rather as manifestations of incomplete acquisition. In particular, these are the cases of interference with particular synonymous Russian verbs without -s j a and overgeneralization of productive meanings of -s j a. Some nonstandard uses of -s j a in the speech of older speakers may be inherited from the local pidgin.
The Tungusic data show a significant prevalence of anticausatives across nonstandard uses ofs j a. This agrees with our expectations on the interference with the anticausative -p in Tungusic. At the same time, deponent verbs do not show any prevalence, as could be expected according to the hypothesis of under-acquisition of the Russian system.
To conclude, we cannot fully explain the picture observed either by direct calquing of the pattern of the indigenous language or by the incomplete acquisition of standard Russian. We are dealing rather with the interaction of both types of factors and probably also with some additional ones.
Funding: This research was funded by RSF grant number 17-18-01649.

Conflicts of Interest:
The authors declare no conflict of interest. (Avrorin 1961

References
Nothing, only your finger bandage is lying there'. (Nganasan corpus  However, in these two examples, there might be alternative explanations. In standard Russian, the relations between sʲa-forms and forms without -s j a are not always so regular, as in reflexive, reciprocal, anticausative, and passive uses. For instance, the relation between Russian verbs val j at j 'to drag' and val j at j -s j a 'to lie' is not transparent. The omission of -s j a in (12) can be explained by an underacquisition of such irregular relations. Moreover, there is a synonymous Russian verb lezhatʲ 'lie'.
On the other hand, there are cases where the omission cannot be explained as a calque.
In (14), the verb ispugat j -s j a is used without -s j a, although the corresponding verb requires the reflexive marker in Enets (15) It seems that there is no equivalent Russian verb without -s j a that could have affected this use. The semantic relation between ispugat j and ispugat j -s j a is regular as well (anticausative). Presumably, this omission could be explained by a general tendency to omit it, which may have been inherited from the local Pidgin variety. Example (14) was taken from a text of the oldest speaker of Forest Enets (1910 year of birth). He probably knew Taimyr Pidgin Russian, or Govorka, which is now extinct (on Govorka see (Stern 2005)). In the basilect of this pidgin, -s j a is indeed regularly omitted (Urmanchieva 2010, p. 199).
The same explanation can be proposed for (16) from Nganasan Russian. This example cannot be attributed to the Nganasan influence since in Nganasan, inchoative verbs have the reflexive conjugation (Tereschenko 1979, p. 195). Example (16) comes from a speaker of an older generation (1923), who is regarded as a mesolect speaker of Govorka 4 .

16.
purge načina-l-sʲa sovsem blizzard begin-PST.SG.M-SJA entirely 'A raging blizzard started'. (L1 Nganasan) Table 2 shows that both in the Tungusic sample and in the Samoyedic one, the number of omissions is higher than the number of overuses. It seems that there is no equivalent Russian verb with The semantic relation between ispugat j and ispugat j -s j a is reg this omission could be explained by a general tendency to from the local Pidgin variety. Example (14) was taken from a (1910 year of birth). He probably knew Taimyr Pidgin Russ Govorka see (Stern 2005)). In the basilect of this pidgin, -s j a i 2010, p. 199).

Quantitative Data
The same explanation can be proposed for (16) from N attributed to the Nganasan influence since in Nganasa conjugation (Tereschenko 1979, p. 195 Table 2 shows that both in the Tungusic sample and omissions is higher than the number of overuses.   However, in these two examples, there might be alternative explanations. In standard Russian, the relations between sʲa-forms and forms without -s j a are not always so regular, as in reflexive, reciprocal, anticausative, and passive uses. For instance, the relation between Russian verbs val j at j 'to drag' and val j at j -s j a 'to lie' is not transparent. The omission of -s j a in (12) can be explained by an underacquisition of such irregular relations. Moreover, there is a synonymous Russian verb lezhatʲ 'lie'.

Quantitative Data
On the other hand, there are cases where the omission cannot be explained as a calque.
In (14), the verb ispugat j -s j a is used without -s j a, although the corresponding verb requires the reflexive marker in Enets (15) It seems that there is no equivalent Russian verb without -s j a that could have affected this use. The semantic relation between ispugat j and ispugat j -s j a is regular as well (anticausative). Presumably, this omission could be explained by a general tendency to omit it, which may have been inherited from the local Pidgin variety. Example (14) was taken from a text of the oldest speaker of Forest Enets (1910 year of birth). He probably knew Taimyr Pidgin Russian, or Govorka, which is now extinct (on Govorka see (Stern 2005)). In the basilect of this pidgin, -s j a is indeed regularly omitted (Urmanchieva 2010, p. 199).
The same explanation can be proposed for (16) from Nganasan Russian. This example cannot be attributed to the Nganasan influence since in Nganasan, inchoative verbs have the reflexive conjugation (Tereschenko 1979, p. 195). Example (16) comes from a speaker of an older generation (1923), who is regarded as a mesolect speaker of Govorka 4 .
16. purge načina-l-sʲa sovsem blizzard begin-PST.SG.M-SJA entirely 'A raging blizzard started'. (L1 Nganasan) Table 2 shows that both in the Tungusic sample and in the Samoyedic one, the number of omissions is higher than the number of overuses. However, in these two examples, there might be alternative explanations. In standard Russian, the relations between sʲa-forms and forms without -s j a are not always so regular, as in reflexive, reciprocal, anticausative, and passive uses. For instance, the relation between Russian verbs val j at j 'to drag' and val j at j -s j a 'to lie' is not transparent. The omission of -s j a in (12) can be explained by an underacquisition of such irregular relations. Moreover, there is a synonymous Russian verb lezhatʲ 'lie'.

Quantitative Data
On the other hand, there are cases where the omission cannot be explained as a calque.
In (14), the verb ispugat j -s j a is used without -s j a, although the corresponding verb requires the reflexive marker in Enets (15) It seems that there is no equivalent Russian verb without -s j a that could have affected this use. The semantic relation between ispugat j and ispugat j -s j a is regular as well (anticausative). Presumably, this omission could be explained by a general tendency to omit it, which may have been inherited from the local Pidgin variety. Example (14) was taken from a text of the oldest speaker of Forest Enets (1910 year of birth). He probably knew Taimyr Pidgin Russian, or Govorka, which is now extinct (on Govorka see (Stern 2005)). In the basilect of this pidgin, -s j a is indeed regularly omitted (Urmanchieva 2010, p. 199).
The same explanation can be proposed for (16) from Nganasan Russian. This example cannot be attributed to the Nganasan influence since in Nganasan, inchoative verbs have the reflexive conjugation (Tereschenko 1979, p. 195). Example (16) comes from a speaker of an older generation (1923), who is regarded as a mesolect speaker of Govorka 4 .
16. purge načina-l-sʲa sovsem blizzard begin-PST.SG.M-SJA entirely 'A raging blizzard started'. (L1 Nganasan) Table 2 shows that both in the Tungusic sample and in the Samoyedic one, the number of omissions is higher than the number of overuses. It seems that there is no equivalent Russian verb with The semantic relation between ispugat j and ispugat j -s j a is reg this omission could be explained by a general tendency to from the local Pidgin variety. Example (14) was taken from a (1910 year of birth). He probably knew Taimyr Pidgin Russ Govorka see (Stern 2005)). In the basilect of this pidgin, -s j a i 2010, p. 199).

Quantitative Data
The same explanation can be proposed for (16) from N attributed to the Nganasan influence since in Nganasa conjugation (Tereschenko 1979, p. 195). Example (16) come (1923), who is regarded as a mesolect speaker of Govorka 4 .
16. purge načina-l-sʲa blizzard begin-PST. 'A raging blizzard started'. (L1 Nganasan) Table 2 shows that both in the Tungusic sample and omissions is higher than the number of overuses.   However, in these two examples, there might be alternative explanations. In standard Russian, the relations between sʲa-forms and forms without -s j a are not always so regular, as in reflexive, reciprocal, anticausative, and passive uses. For instance, the relation between Russian verbs val j at j 'to drag' and val j at j -s j a 'to lie' is not transparent. The omission of -s j a in (12) can be explained by an underacquisition of such irregular relations. Moreover, there is a synonymous Russian verb lezhatʲ 'lie'.

Quantitative Data
On the other hand, there are cases where the omission cannot be explained as a calque.
In (14), the verb ispugat j -s j a is used without -s j a, although the corresponding verb requires the reflexive marker in Enets (15) It seems that there is no equivalent Russian verb without -s j a that could have affected this use. The semantic relation between ispugat j and ispugat j -s j a is regular as well (anticausative). Presumably, this omission could be explained by a general tendency to omit it, which may have been inherited from the local Pidgin variety. Example (14) was taken from a text of the oldest speaker of Forest Enets (1910 year of birth). He probably knew Taimyr Pidgin Russian, or Govorka, which is now extinct (on Govorka see (Stern 2005)). In the basilect of this pidgin, -s j a is indeed regularly omitted (Urmanchieva 2010, p. 199).
The same explanation can be proposed for (16) from Nganasan Russian. This example cannot be attributed to the Nganasan influence since in Nganasan, inchoative verbs have the reflexive conjugation (Tereschenko 1979, p. 195). Example (16) comes from a speaker of an older generation (1923), who is regarded as a mesolect speaker of Govorka 4 .
16. purge načina-l-sʲa sovsem blizzard begin-PST.SG.M-SJA entirely 'A raging blizzard started'. (L1 Nganasan) Table 2 shows that both in the Tungusic sample and in the Samoyedic one, the number of omissions is higher than the number of overuses.   However, in these two examples, there might be alternative explanations. In standard Russian, the relations between sʲa-forms and forms without -s j a are not always so regular, as in reflexive, reciprocal, anticausative, and passive uses. For instance, the relation between Russian verbs val j at j 'to drag' and val j at j -s j a 'to lie' is not transparent. The omission of -s j a in (12) can be explained by an underacquisition of such irregular relations. Moreover, there is a synonymous Russian verb lezhatʲ 'lie'.

Quantitative Data
On the other hand, there are cases where the omission cannot be explained as a calque.
In (14), the verb ispugat j -s j a is used without -s j a, although the corresponding verb requires the reflexive marker in Enets (15) It seems that there is no equivalent Russian verb without -s j a that could have affected this use. The semantic relation between ispugat j and ispugat j -s j a is regular as well (anticausative). Presumably, this omission could be explained by a general tendency to omit it, which may have been inherited from the local Pidgin variety. Example (14) was taken from a text of the oldest speaker of Forest Enets (1910 year of birth). He probably knew Taimyr Pidgin Russian, or Govorka, which is now extinct (on Govorka see (Stern 2005)). In the basilect of this pidgin, -s j a is indeed regularly omitted (Urmanchieva 2010, p. 199).
The same explanation can be proposed for (16) from Nganasan Russian. This example cannot be attributed to the Nganasan influence since in Nganasan, inchoative verbs have the reflexive conjugation (Tereschenko 1979, p. 195). Example (16) comes from a speaker of an older generation (1923), who is regarded as a mesolect speaker of Govorka 4 .
16. purge načina-l-sʲa sovsem blizzard begin-PST.SG.M-SJA entirely 'A raging blizzard started'. (L1 Nganasan) Table 2 shows that both in the Tungusic sample and in the Samoyedic one, the number of omissions is higher than the number of overuses. 13. dʼaŋku taa-ni-ə dʼüðü-tə NEG.EX that.remote-LOC.PRON-ADJZ hand-GEN.SG.2SG sʼügümü-ə-dʼəə-raa takəə dʼübə-i-ti bandage-ADJZ-ANT-LIM that.remote throw-DRV[STAT]-PRS nʼülʼi͡ a-jtʼi-tɨ lie.down.straight-DRV-PRS 'Nothing, only your finger bandage is lying there'. (Nganasan corpus  However, in these two examples, there might be alternative explanations. In standard Russian, the relations between sʲa-forms and forms without -s j a are not always so regular, as in reflexive, reciprocal, anticausative, and passive uses. For instance, the relation between Russian verbs val j at j 'to drag' and val j at j -s j a 'to lie' is not transparent. The omission of -s j a in (12) can be explained by an underacquisition of such irregular relations. Moreover, there is a synonymous Russian verb lezhatʲ 'lie'.

Quantitative Data
On the other hand, there are cases where the omission cannot be explained as a calque.
In (14), the verb ispugat j -s j a is used without -s j a, although the corresponding verb requires the reflexive marker in Enets (15) It seems that there is no equivalent Russian verb without -s j a that could have affected this use. The semantic relation between ispugat j and ispugat j -s j a is regular as well (anticausative). Presumably, this omission could be explained by a general tendency to omit it, which may have been inherited from the local Pidgin variety. Example (14) was taken from a text of the oldest speaker of Forest Enets (1910 year of birth). He probably knew Taimyr Pidgin Russian, or Govorka, which is now extinct (on Govorka see (Stern 2005)). In the basilect of this pidgin, -s j a is indeed regularly omitted (Urmanchieva 2010, p. 199).
The same explanation can be proposed for (16) from Nganasan Russian. This example cannot be attributed to the Nganasan influence since in Nganasan, inchoative verbs have the reflexive conjugation (Tereschenko 1979, p. 195). Example (16) comes from a speaker of an older generation (1923), who is regarded as a mesolect speaker of Govorka 4 .

16.
purge načina-l-sʲa sovsem blizzard begin-PST.SG.M-SJA entirely 'A raging blizzard started'. (L1 Nganasan) Table 2 shows that both in the Tungusic sample and in the Samoyedic one, the number of omissions is higher than the number of overuses.   However, in these two examples, there might be alternative explanations. In standard Russian, the relations between s j a-forms and forms without -s j a are not always so regular, as in reflexive, reciprocal, anticausative, and passive uses. For instance, the relation between Russian verbs val j at j 'to drag' and val j at j -s j a 'to lie' is not transparent. The omission of -s j a in (12) can be explained by an under-acquisition of such irregular relations. Moreover, there is a synonymous Russian verb lezhat j 'lie'.

Quantitative Data
On the other hand, there are cases where the omission cannot be explained as a calque.
In (14), the verb ispugat j -s j a is used without -s j a, although the corresponding verb requires the reflexive marker in Enets (15) It seems that there is no equivalent Russian verb without -s j a that could have affected this use. The semantic relation between ispugat j and ispugat j -s j a is regular as well (anticausative). Presumably, this omission could be explained by a general tendency to omit it, which may have been inherited from the local Pidgin variety. Example (14) was taken from a text of the oldest speaker of Forest Enets (1910 year of birth). He probably knew Taimyr Pidgin Russian, or Govorka, which is now extinct (on Govorka see (Stern 2005). In the basilect of this pidgin, -s j a is indeed regularly omitted (Urmanchieva 2010, p. 199).
The same explanation can be proposed for (16) from Nganasan Russian. This example cannot be attributed to the Nganasan influence since in Nganasan, inchoative verbs have the reflexive conjugation (Tereschenko 1979, p. 195). Example (16) comes from a speaker of an older generation (1923), who is regarded as a mesolect speaker of Govorka 4 .

purge načina-l
Languages 2019, 4, x FOR PEER REVIEW 8 of differs from standard Russian. Moreover, sometimes we witness a variation: -sʲa can be omitted an used correctly within one paragraph or even within one sentence, as in (17) We do not observe notable differences between the Samoyedic and Tungusic data. Howeve this might be partly explained by the extremely small number of nonstandard -s j a uses in th Samoyedic sample.
We divided all nonstandard uses of -s j a into two groups: omissions (the unexpected absence o -s j a) and overuses (the unexpected presence of -s j a). Both in the Tungusic text sample and in th Samoyedic one, omissions were more frequent than overuses. This generally agrees with ou expectations on the influence of the indigenous system. The prevalence of overuses indeed is no expected, unless the correlate of -s j a in the source language was much more productive. This was no the case either in Tungusic or in Samoyedic.
However, if we exclude deponent verbs, for which omission is logically the only option, an irregular s j a-derivates, for which overuses are not attested either, overuses, in contrast, become eve more frequent than omissions.
Not all nonstandard uses of -s j a are caused by structural borrowing. There are even more case that can be interpreted rather as manifestations of incomplete acquisition. In particular, these are th  Table 2 shows that both in the Tungusic sample and in the Samoyedic one, the number of omissions is higher than the number of overuses.  Table 3 shows the distribution of overuses and omissions across different meanings typical of -s j a (on the data of Tungusic subcorpus 5 ). A significantly higher rate of nonstandard uses compared to standard ones is attested for anticausative s j a-verbs (such as lomat jlomat j -s j a 'break (intransitive)-break (transitive)') 6 . Deponent verbs and verbs with an irregular semantic relation to the base verb, which amounts to the majority of s j a-verbs used in a nonstandard way, do not deviate significantly from that of other semantic classes of s j a-verbs.  Table 4 shows the correlation between the meaning of -s j a and the type of nonstandard use. As expected, only omissions are attested across deponent verbs and verbs with irregular semantic relations between the base verb and the derived form and all overuses belong to the productive meanings of -s j a. Moreover, across productive meanings, the above-mentioned asymmetry between omissions and overuses is not attested. Table 4. Different meanings of -s j a: omission vs. overuse (Tungusic and Samoyedic) 8 .

Omission Overuse
Deponent and irregular 27 0 productive meanings 19 25 Table 5 demonstrates the distribution of overuses of -s j a motivated by different factors. Clear cases of structural borrowing are rarer than cases of incomplete acquisition (the interference with a synonymous verb without -s j a and overgeneralization of productive meanings of -s j a).

Discussion
Thus, we have analyzed the nonstandard uses of the reflexive suffix -s j a in the Russian speech of bilingual speakers of indigenous languages of Siberia, namely Samoyedic and Tungusic languages. Such uses are quite infrequent in the text sample. Since there are many more uses of -s j a that follow the rules of standard Russian, the uses observed in the data do not form a consistent system that differs from standard Russian. Moreover, sometimes we witness a variation: -s j a can be omitted and used correctly within one paragraph or even within one sentence, as in (17) differs from standard Russian. Moreover, sometimes we witness a variation: -sʲa can be omitted and used correctly within one paragraph or even within one sentence, as in (17) We do not observe notable differences between the Samoyedic and Tungusic data. However, this might be partly explained by the extremely small number of nonstandard -s j a uses in the Samoyedic sample.
We divided all nonstandard uses of -s j a into two groups: omissions (the unexpected absence of -s j a) and overuses (the unexpected presence of -s j a). Both in the Tungusic text sample and in the Samoyedic one, omissions were more frequent than overuses. This generally agrees with our expectations on the influence of the indigenous system. The prevalence of overuses indeed is not expected, unless the correlate of -s j a in the source language was much more productive. This was not the case either in Tungusic or in Samoyedic.
However, if we exclude deponent verbs, for which omission is logically the only option, and irregular s j a-derivates, for which overuses are not attested either, overuses, in contrast, become even more frequent than omissions.
Not all nonstandard uses of -s j a are caused by structural borrowing. There are even more cases that can be interpreted rather as manifestations of incomplete acquisition. In particular, these are the cases of interference with particular synonymous Russian verbs without -s j a and overgeneralization of productive meanings of -s j a. Some nonstandard uses of -s j a in the speech of older speakers may be inherited from the local pidgin.
The Tungusic data show a significant prevalence of anticausatives across nonstandard uses ofs j a. This agrees with our expectations on the interference with the anticausative -p in Tungusic. At the same time, deponent verbs do not show any prevalence, as could be expected according to the hypothesis of under-acquisition of the Russian system.
To conclude, we cannot fully explain the picture observed either by direct calquing of the pattern of the indigenous language or by the incomplete acquisition of standard Russian. We are dealing rather with the interaction of both types of factors and probably also with some additional ones. We do not observe notable differences between the Samoyedic and Tungusic data. However, this might be partly explained by the extremely small number of nonstandard -s j a uses in the Samoyedic sample.
We divided all nonstandard uses of -s j a into two groups: omissions (the unexpected absence of -s j a) and overuses (the unexpected presence of -s j a). Both in the Tungusic text sample and in the Samoyedic one, omissions were more frequent than overuses. This generally agrees with our expectations on the influence of the indigenous system. The prevalence of overuses indeed is not expected, unless the correlate of -s j a in the source language was much more productive. This was not the case either in Tungusic or in Samoyedic.
However, if we exclude deponent verbs, for which omission is logically the only option, and irregular s j a-derivates, for which overuses are not attested either, overuses, in contrast, become even more frequent than omissions.
Not all nonstandard uses of -s j a are caused by structural borrowing. There are even more cases that can be interpreted rather as manifestations of incomplete acquisition. In particular, these are the cases of interference with particular synonymous Russian verbs without -s j a and overgeneralization of productive meanings of -s j a. Some nonstandard uses of -s j a in the speech of older speakers may be inherited from the local pidgin.
The Tungusic data show a significant prevalence of anticausatives across nonstandard uses of -s j a. This agrees with our expectations on the interference with the anticausative -p in Tungusic. At the same time, deponent verbs do not show any prevalence, as could be expected according to the hypothesis of under-acquisition of the Russian system.
To conclude, we cannot fully explain the picture observed either by direct calquing of the pattern of the indigenous language or by the incomplete acquisition of standard Russian. We are dealing rather with the interaction of both types of factors and probably also with some additional ones.