The p ervasiveness of l anguage c ontact: Evidence from n egative e xistentials in Romeyka/Turkish c ode-s witching

. This paper investigates the morpho-syntactic features of language contact in the endangered Greek dialect Romeyka with Turkish. We analyze the use of the borrowed negative existential jok to (a) determine its role in Romeyka’s negation patterns (b) examine the effects of contact in Romeyka through cross-linguistic comparisons of jok with Turkish and forms of the dialect as spoken in Greece and (c) apply the identified grammatical patterns of jok to Myers-Scotton’s linguistic explanations for the code switching phenomena in the Matrix Language Turnover Hypothesis. The analysis demonstrates the pervasive influence of Turkish on the morpho-syntax of Romeyka through the incorporation of Turkish grammatical structures. We observe changes in the fundamental predicate grammar that are aligned with Turkish and that are inconsistent with Pontic’s existential constructions where the verb indicating existence is used. The patterns of contact confirm the Matrix Language hypothesis and provide evidence that indicate that Romeyka may be undergoing language turnover. Our findings are relevant to further understanding code switching among speakers of minority languages and assessing the vitality of Romeyka in Turkey.

comparison of the jok, we examine the pervasiveness and characteristics of CS in the negator.We study the result of interaction with Turkish, examining the properties of jok in relationship with those of the Turkish negative existential yok.We include comparisons to constructions in Pontic Greek as spoken in Greece to isolate the role of contact.In doing so, we determine evidence of changes in linguistic matter and linguistic patterns due to contact.Finally, the identified grammatical patterns will be used to apply the Matrix Language hypothesis, the central tenet of Myers-Scotton's (1993) linguistic accounts of CS, to observe the extent to which contact has affected Romeyka and understand the trajectory of whether Romeyka based on the presence of language turnover.

Romeyka.
Romeyka is a form of the Greek dialect Pontic Greek spoken in regions of North-Eastern Turkey by an estimated 4000-8000 remaining speakers, making it severely endangered (Sitaridou 2013).Romeyka exhibits rich linguistics that are valuable to research, as it retains many archaic features.The language was historically spoken outside Greece until the Treaty of Lausanne between Greece andTurkey in 1923 (Kaltsa andSitaridou 2010).As a result, many Romeyka speakers of Turkey had to emigrate to Greece.However, there are some remaining Muslim Romeyka speakers of Turkey that have been preserving their language and identity.Many Romeyka speakers in Turkey are bilingual speakers of Turkish and Romeyka.The close coexistence with Turkish both culturally and individually makes the Turkish influence inevitable.
Figure 1.Romeyka is spoken in North-Eastern Turkey In its social and historical context, using Romeyka in social life has not been easy.Language policy in Turkey has had a role in shaping the social status of Romeyka speakers and has changed over the years.In the 1980's, with the "Law Concerning Publications and Broadcasts in Languages Other Than Turkish" (McGonagle et al. 2003), Turkish was defined as the mother tongue of Turkish citizens, prohibiting the use of other languages as a mother tongue.Further restrictions were set on the language in broadcasts and the media.The latter policy was lifted later in 2002, yet still, minority languages face social prejudice and Turkey's vision for cultural hegemony.Similarly with some other minority languages in Turkey, Romeyka does not have a formal written system, and therefore many speakers use either the Turkish Latin alphabet or very rarely Modern Greek alphabet to write in Romeyka.As a language in contact, it is especially fragile.Through our investigation, our aims are to address important and relevant issues regarding endangered languages and the greater implications of language contact.

Theoretial background.
The theoretical framework which we employ to examine the convergence of the Romeyka language is based on Myers Scotton's (1993) Matrix Language Frame model (MLF).The Matrix Language turnover hypothesis explains the morphosyntactic structures general of bilingual speech.It especially discusses structural borrowing in relation to language attrition, and describes contact through the borrowing of inflections and function words as indicative of a risk for language attrition.
The principles of the MLF expand into two auxiliary theories: the 4-M model and the Abstract Level model.The 4-M model considers the distribution of different types of morphemes in language production.The Abstract Level model accounts for features from different varieties structure utterances in convergence.
In the Abstract Level model, Myers distinguishes classic codeswitching from composite codeswitching.In the former, the morphosyntactic frame is based solely from the Matrix Language (ML).Content morphemes, e.g.nouns, verbs, adjectives, and prepositions, can be inserted from the other participating language, the Embedded Language (EL).However, in composite codeswitching, the morphosyntactic frame is provided by both of the participating languages, resulting in features of some grammatical structures from the EL.
The MLF model's explanations of complex language contact phenomena in bilingual speakers relies on a hierarchy of contact and borrowing.Thomason and Kaufman's (1988) borrowing scale describes the degree of resistance of grammar to borrowing based on their systematically.The scale contains levels of contact ranging from casual to intense and the resulting change in the EL.Here, content words or minor phonological features may be related to less contact than word order changes and major changes in morphosyntax.Matras (2007) proposed another hierarchy, based on an investigation of a sample of 27 languages: nouns, conjunctions > verbs > discourse markers > adjectives > interjections > adverbs > other particles, adpositions > numerals < pronouns > derivational affixes > inflectional affixes.Matras explained that grammatical components considered crucial to meaning, including existentials and negators, are rarely borrowable and pattern with the language of finite predication (Matras 2003).The fundamental importance of such structures in the predication grammar makes them resistant in contact, and evidence of borrowing in this level may indicate heavy contact.
Due to the circumstances of Romeyka and our access to only a relatively small data corpus, a quantitative analysis might not sufficiently reveal significant distribution patterns of grammatical elements in the corpus.Instead, the present study focuses on the morphosyntactic patterns in the Romeyka negation system before evaluating the plausibility of language turnover as per the MLF hypothesis.We narrow our scope of study to analyze the pervasiveness of the contact exhibited by the borrowed negative existential, jok.

Methodology.
4.1.CORPUS.In the research we use the corpus named "Word Order in Western Asia" (henceforth WOWA), working with the Romeyka dataset.The data is open source and available for access online.It is constructed from spoken data of two dialects, Romeyka of Of in Caykara and in Sürmene.The data was collected in 2019 in Turkey by Laurentia Schreiber, consisting of freespoken narratives of daily life and personal anecdotes as well as stories prompted by picture card tasks.Schreiber's original aims for the use of the corpus are outlined by her ongoing work, including studying aspects of the influences of language contact, language shift and internal language on the morphosyntax of Romeyka (Schreiber 2021).In order to address this, Schreiber's goals while creating the corpus were to contrast contact-induced language change and native language attrition in language decline.The corpus texts are transcribed, translated into English, and morphologically glossed to construct the dataset, published in 2021.The formatting of the data is done so that each row provides a token from the text and its context, along with translations.The corpus contains various parameters for each text that include latitude, longitude, document details, token, token context, English translations, animacy, weight, role, and flag.We primarily focus on a subset of the WOWA corpus containing jok and other identified negatives of study.4.2.DATA ANALYSIS.The corpus linguistics approach of our research followed a study of the WOWA corpus.Negatives (apart from jok) are first identified on the basis of a computational analysis.We performed a dependency parsing on the English translated token contexts by utilizing spaCy,1 an open-source software library for advanced natural language processing, a subfield of computer science and linguistics studying the computational analysis of text.SpaCy uses computational linguistics to perform part of speech tagging and identify syntactic dependencies within sentences.The functionality of the tool can be seen as follows, where the spaCy identifies the parts of a sentence and can locate the presence of a negation with the dependency marker, neg:

Figure 2. spaCy's dependency parsing functionality illustrated
The tool provides state-of-the-art speed and accuracy in working with corpora, and enabled us to locate where negation was present in order to determine the negators in the original Romeyka text.We observed the following negative particles in the documents, multiple of which exhibit phonologically conditioned variation (Sitaridou 2014): jo/jog/jok, mi, and tsh/u/udzh/utsh/utshe/uzh. The functionality and usage of each negator in the corpus are discussed.
Utsh and its allomorphic forms, including tsh, u, udzh, utshe, and uzh, occur in sentential negation of declarative sentences.In example (1) we show the use of utsh in a declarative sentence.
(1) Romeyka (Schreiber 2021) Xastas utsh ine.sick NEG be.1SG 'I am not sick.'Utsh negates the verb ime, 'to be'.It acts as the negator for standard negation.Standard negation refers to the negation strategy used in declarative sentences where the predicate is a full lexical verb.
The u form of utsh follows the same pattern as in (1), as shown in (2).
The negator mi appears in imperatives and wishes.Its use in an imperative is presented in example (3).
(3) Romeyka (Schreiber 2021) mi bas sin avlea… I inega ipen adona POSS woman say.PAST.3SGman NEG go to.the.ACC house 'His wife said to him, do not go in front of the house…' Here, the wife tells the man to not go, and mi can be observed negating bas 'go'.Similar to utsh, mi precedes its verbal predicate.
Nonexistence is expressed through the existential negative jok and its forms.Example (4) presents the use of jok to indicate absence.
(4) Romeyka (Schreiber 2021) Jok su Istaboli.NEG.EX in.ACC Istabul '(this) does not exist in Istanbul.'Jok here does not have a verbal predicate and expresses the lack of something without an explicitly stated subject.A non-referential subject is a typical characteristic of many existential sentences.
The morpheme jok is borrowed from the Turkish counterpart yok, which is shown with the same function, a negative existential, in example (5).
(5) Turkish Ev-de yok.home-LOC NEG.EX '(this) does not exist in Istanbul.' Here, the existential properties of the Turkish negator yok can be observed, as it expresses the absence of an entity at home, a prepositional phrase.The phrase is used as a response to a question, denying existence.
For the purposes of our study, we applied manipulations and further additions to the corpus in order to align with our research focus and study of jok.In order to establish the role of jok in Romeyka negation, we additionally compared the contexts of the negators with respect to the sentence types the negators are present in and their word order.This comparative analysis methodology enabled us to compare the negation system of Romeyka and make contrastive observations on jok.These establish a background overview of the negation structures found in the WOWA corpus.
After establishing the role of jok in the Romeyka negation system, we analyze case by case its utterances to determine the extent of contact which occurs in its use.In order to do so we look at the borrowing and the presence of structural changes, comparing instances to Turkish and the Pontic Greek of Greece to put together the influence of contact.As Pontic Greek as spoken in Greece and Romeyka are of the same dialect, comparisons based on their locations and contact with Turkish provide a direct observation of the changes present.
These steps enable us to interpret our findings through comparisons to the Matrix Language Turnover hypothesis and identify the viability and oath for preservation of Romeyka.

THE ROLE OF JOK.
In this section, we determine the role of jok in the system of Romeyka, aiming to answer whether it is fundamentally different from the other negators and the differences in their uses.We provide an outline of the distribution of the negation patterns as they appear in the WOWA corpus.

Total text segments 502
Total We examine the sentence types in which the negators appear.We observe declarative, imperative, and counterfactual phrases in our corpus.As aforementioned, utsh appears in declaratives and mi in imperatives and at times in counterfactuals.Jok, under consideration of sentence structure, is most similar to utsh, observed in declaratives.Here, it is unclear whether jok serves a unique purpose in the system.
It is important to note that existentials often occur as declaratives.They are typically distinguished with the characteristics of a non-referential subject, distinctive word order, and/or special agreement or no agreement between subject and predicate.
Word order of the negators in relation to the token which they negate, expressed as a difference of the negator's position in a phrase and the position of what is negated, displays a variation between the Utsh precedes what it negates, as shown in ( 6). ( 6) Romeyka (Schreiber 2021) Ado utsh eksero.that NEG know.1SG'That I do not know' In ( 6), utsh negates the verbal predicate eksero, 'to know', appearing before the verb to indicate its negation.This order is the structure for standard negation in Romeyka.
Jok shows a constructional difference in its negation strategy from standard negation with utsh.
It does not precede nor have verbal predicates and instead shows its own verbal properties.
(7) Romeyka (Schreiber 2021) Jane hadha so mehele jok.well here in.the.ACC neighborhood NEG.EX 'Well, in this neighborhood there was no (school).'(8) Jok su Istaboli.NEG.EX in.ACC Istabul '(this) does not exist in Istanbul.'In (7), jok is used in a locative existential sentence, following the negated locative predicate.Furthermore, the subject is not explicitly stated in the sentence, as the speaker does not reference the school.This characteristic is also observed in (8).These existential sentences are context-dependent, getting meaning from their preceding sentences.(7), however, shows the negative, jok, being used before the locative which it modifies, Istanboli, 'Istanbul'.This version of the word order appears in the corpus only in constructions following this same pattern, while the first word order is most common in the use of jok.
From these observations, we establish that jok shows distinctions from the rest of the Romeyka negation system.It exhibits a complete constructional and prototypical difference from standard negation as it appears in our corpus.

FEATURES OF LANGUAGE CONTACT.
Jok of Romeyka is borrowed from the Turkish negative existential yok.In Turkish, yok is a strong negator serving the purpose of indicating absence or non-existence.We compare the grammar and use of jok in Turkish and Romeyka to observe the extent of the contact in Romeyka's negation.From our corpus, we study the utterances of jok to determine similarities and understand the influence of Turkish.
Jok is observed in the corpus in two types of existential sentences: locative existential sentences and bare existential sentences.Locative types of existential sentences are exemplified as follows: (9) Romeyka (Schreiber 2021) a. Jok su Istaboli.(11b) presents a locative existential sentence in Turkish with a pivot Istanbul-da 'in Istanbul' marked with a locative case ending and the existential negator yok.It is dependent on the question in (11a), where the existential var is used to ask the existence of the presented subject.The use of yok in (11b) and (11c) aligns with the use of jok in Romeyka as in (9).In (11d), the bare existential mirrors (10), with a pivot and existential.Turkish also has possessive existential constructions, however, the size of our data did not provide sufficient evidence of whether these occur in Romeyka as well.
Another use of jok borrows the context of the use of yok in Turkish for cardinality.
(32) Romeyka (Schreiber 2021) …pek jok... much NEG.EX '…not much / there isn't much...' (43) Turkish pek yok.much NEG.EX '…not much / there isn't much...' In ( 12), jok is used with the Turkish indefinite quantifier pek 'much/quite'.This is a cooccurance common in Turkish, as in ( 13), and is used as a term to indicate paucity or lack of abundance.It is in contrast from the use of an existential to deny absolutely.Borrowing the surrounding context of jok inserts further Turkish components into Romeyka.
The types of constructions that jok and yok appear in are the same and they serve the same semantic and discourse functions..It is interesting to note that existential constructions with jok defy the dominant grammar patterns of Romeyka and have various properties that follow Turkish.The existential predicates of jok are verbal, acting as a verb without a copula.
The use of jok in contrast to its corresponding copular sentence using the negator utsh is shown in ( 14). ( 14) Romeyka (Schreiber 2021) a. ...hadha so mehele jok here in.ACC neighborhood NEG.EX '...here in the neighborhood there was no school.'b. S emetero to mehele ektebi utsh edon.in.ACC POSS.3SG the.ACC neighborhood.ACC school NEG be 'In our neighborhood there was no school.'Existential sentences state the existence of an object, and typically take from the characteristics of having a non-referential subject or a word order that differs from the dominant word orders in the language.Negative existentials deny existence absolutely and themselves describe the lack of existence, distinguishing it from negation marker (Seydi 2020).In (14a), jok acts as a negation of a presumed existence of the school with a locative predicate of the neighborhood.The subject is omitted.It displays a constructional difference from expressions used for standard negation as in (14b).In (14b) the predicate edon 'to be' is a full lexical verb that is modified by the negator utsh.(14a) and (14b) differ in their composition and syntactic structure.Whereas in (14a), utsh negates edon, (14b) shows no verbal predicate and plainly functions to communicate absence itself.In example (14b), the existential construction patterns of Turkish are followed, which uses yok primarily with locative predicates and displays a similar syntactic order.
(55) Turkish Bura-da mahalle-de okul yok.here-LOC neighborhood-LOC school NEG.EX 'Here in the neighborhood there was no school.' In the above example, Turkish's use of yok as a negative existential is shown.Comparing (14a) and ( 15) we observe the lexical borrowing of Turkish yok as Romeyka jok, both serving the same function of indicating non-existence.In contrast of the two compositions to (14b), the deviation, grammatically, from the standard means of negation in Romeyka can be observed.( 14a) and ( 15) share the lack of a verb which the negator modifies.Through contact, not only is the morpheme borrowed, but despite its inconsistency with Romeyka's fundamental grammar, grammar is also maintained.
The absence of copula in constructions with jok is particularly notable in bare existentials.Sansa Tura (1986) argues that unlike locative existentials, bare existentials do not have a corresponding copular sentence.This quality means the presence of bare existentials in Romeyka is unique to constructions with the borrowed jok.
The influence of Turkish on the grammatical qualities of Romeyka are further illustrated by ( 16), which presents the existential patterns of Pontic as spoken in Greece.As Romeyka is a form of Pontic Greek, observing the characteristics in the form spoken outside of Turkey provides a comparison of the impact of contact with Turkish on the language.We compare these two forms of the same dialect to demonstrate the role of contact with Turkish.In Pontic Greek, existentials are constructed using the morpheme ine (Sitaridou 2014).It is illustrated in ( 16): (66) Pontic Greek (Sitaridou 2014) Xorafae kh' ine.fields NEG exist.3PL 'There are no fields.'In ( 16), the existential ine is modified by the negator kh' to indicate the lack of existence of the coda xorafae 'fields'.The behavior of jok differs from this pattern, as the negator kh' negates the verb for existence ine, while, in the case of jok, a proposition about the lack of presence is expressed in itself.Jok does not modify a verb like kh' in ( 16).The use of jok is inconsistent with the verbal predicates and use of 'to be' otherwise observed in Romeyka negation.
Ine is also still used in Romeyka existentials, shown below in ( 17), but it does not appear in the use of jok.The use of kh' is not found in our corpus.
(77) Romeyka (Schreiber 2021) Opsara ine.Fish exist.3PL'There are fish.'Jok follows different patterns in its use that reflect those of Turkish and indicate the reach of contact beyond borrowing of the morpheme into the negation system, but also transferring the grammar of its use in a manner that does not fit with its grammar.Constructions with jok are distinct from other existential constructions in Romeyka.

EVIDENCE OF LANGUAGE TURNOVER
BY APPLICATION OF THE MATRIX FRAME MODEL.As an endangered language with a declining number of speakers, considering the vitality of Romeyka is important to ensure its survival.The pervasiveness of language contact in our study of jok provides insight into the causes and implications of language contact in Romeyka.To explain the trajectory and risk of Romeyka, we interpret our findings to apply to the Matrix Language (ML) Turnover hypothesis from the scope of its negation and borrowing of jok.Myers-Scotton (1993) explains that convergence involves a change in the morpho-syntactic frame of language production.The MLF model distinguishes between the borrowing of content and system morphemes in their fundamental importance.Content morphemes, e.g.nouns, verbs, adjectives, and some prepositions, express semantic and pragmatic meaning in communication.System morphemes, e.g.function words and inflections, express the relation between content morphemes and build grammatical frames.Myers-Scotton characterizes this latter structural borrowing and borrowing of function words to be indicative of leading towards language convergence.
Jok demonstrates codeswitching in system morphemes, as in ( 18). ( 18) Romeyka (Schreiber 2021) Jok su Istaboli.NEG.EX in.ACC Istabul '(this) does not exist in Istanbul.' The content morphemes in (18) are su Istanboli, while the system morpheme is jok, which modifies the meaning of the sentence but does not supply a root meaning.Matras' (2003) hierarchy explains that existentials and negators are grammatical components considered crucial to meaning.The fundamental importance of such structures in the predication grammar makes them resistant in contact, and evidence of borrowing in this level through jok supports the ML hypothesis.
Further observations, shown below, in the use of jok can be explained by the ML hypothesis.
(  19), jok does not negate with a verb and contains no copular predicate that it modifies, while its negation constructions with utsh relies on some homonym of 'to be'.The MLF model's blocking hypothesis explains that any EL content morpheme which is not congruent with the ML is blocked by a filter.The blocking of verbal phrases in Romeyka by jok in order to follow the grammar of Turkish's yok indicates the stages predicted in the ML hypothesis.
Additionally, the implications of the MLF model present the unit hypothesis, which states that multimorphemic EL insertions are unit inserted into a ML clausal frame.These EL "islands" are often adverbial phrases (Myers-Scotton 2002).Myers-Scotton accounts for double morphology in her model.
The linguistic patterns in Romeyka's fundamental grammar and occurrences of bilingual speech in borrowed morphemes can be explained by the ML hypothesis.The fundamental importance of negatives in the predication grammar and the pervasiveness they denote through their contact suggest that Romeyka shows indications of approaching language turnover through contact with Turkish.
In many converging languages, a minority culture is viewed by a dominant culture as a marker of cultural differentness, and it has been a primary cause (Myers-Scotton 2012).The case of Romeyka presents a need for promotion of diversity in order to counter this, as language policy in Turkey has established the dominant language, Turkish, as being symbolic of citizenship of the nation.Our research reveals effective paths for preservation efforts and explicates the scale of the influence that contact has on Romeyka's endangerment.

Conclusion.
This research was aimed to examine the patterns of the borrowed negative existential jok in Romeyka.In our work, we contribute to a documentation of Romeyka's linguistics, provide a study of the changes in linguistic patterns through contact in Romeyka, and determine the impact of such influence in the language's viability and convergence.The linguistic analysis indicates that Turkish significantly influences the morpho-syntactic frame of the constituent.Our examples demonstrate the fundamental role of negative existentials in the predication grammar of Romeyka and the significant absence of verbal predicates with jok.In fact, the use of jok aligned with the use of yok in Turkish and diverges from negatives in Romeyka and other forms of Pontic Greek outside of Turkish influence.Based on the results, there is considerable verifica-tion of Myers-Scotton's Matrix Language Turnover Hypothesis in the corpus, and despite only having looked at negatives, the observations indicate evidence of Romeyka being in the process of a turnover, which suggests that there is a phase in which a language becomes a composite.Our study was limited by our scope of the negative system of Romeyka as well as the size of our data samples.To better understand the relation between the two languages, this study can be furthered by extending the study to forms of CS outside of jok and studying a larger sample of data.We also note the narrative nature of the spoken data, as negatives may be less frequently pre-sent.
As Romeyka's contact with Turkish affects its vitality into the future, understanding the degrees of influence become valuable to the language's preservation and the sustenance of diverse languages.Our research provides insight into the social context of the reasons behind language contact and what it means for the trajectory of endangered and minority languages.Our studies are relevant in future studies on the role specifically of language policy on the social attitudes of speakers towards preserving their language and examinations to the best methods to countering language endangerment for languages who are in unstable contact.Although there has been some research on Romeyka and its grammatical structure, its study has been limited.This study opens more exciting research questions waiting to be addressed.

Table 1 .
A summary of the corpus and the sentence types in which the negators appear within it Such constructions do not have existential objects.The functions of jok are in line with those of yok in Turkish.Yok appears in the same types of existential constructions, as shown in (11).
'here there is no such thing.'Ascanbeseen in (9a), there is a locative phrase or coda su Istanboli 'in Istanbul' and the existential negator jok 'not exist'.Interestingly, there is no noun phrase or pivot in the sentence, and instead, this is mentioned in the speaker's previous utterance.The examples act as a response to a question to deny existence.The existential is thus dependent on the context.These patterns are consistent in (9b), (9c), and (9d).The function of locative existentials in Romeyka is to indicate the lack or absence of something at a particular location.Another type of existential is the bare existential construction.Unlike locative 'There is nothing.'Asobserved in (10), bare existential sentences contain a pivot noun phrase and the existential predicate jok 'not exist'.