Recycling a Mixed Language: Posha in Turkey

Abstract

We provide a sketch grammar of a new bilingual mixed language based on data gathered through interaction with its last native speakers. The language, which we call Posha of Çankırı, is spoken in central Turkey. The source languages are Turkish and Lomavren, another bilingual mixed language for which the source languages are Armenian and some Central Indo-Aryan varieties. In Posha of Çankırı, the mixing happens in the nominal morphology and in the lexicon while the verbal roots and verbal morphology are entirely from the ancestral language, Lomavren, albeit with certain minor changes. The Indo-Aryan layer of vocabulary is rather thin and the Indo-Aryan retentions in grammar can only be speculated. We show that the emergence of Posha of Çankırı has been initiated by language shift, but that its ultimate defining characteristic is L2 insertions into (some distorted version of) the L1. The study contributes to the documentation of lesser known new varieties and touches upon topics such as the mechanism involved in the emergence of bilingual mixed languages.

1. Introduction

Bilingual mixed languages, i.e., languages in which different sub-systems derive from different languages (Thomason and Kaufman 1988, p. 12), have constituted an important research domain in contact linguistics since at least Heine (1969; cited in Mous 2003). One major point of discussion revolves around what contact mechanisms lead to their formation (e.g., language shift, cf. Matras 2000 for an overview; relexification, Muysken 1981; lexical re-orientation, Matras 2003 a.o.; code-switching, Mous 2003; Muysken 1997, 2007; O’Shannessy 2012). Mixed languages clearly do not form a homogeneous category, neither with respect to the mechanisms leading to their emergence nor with respect to the sociolinguistic conditions in which these mechanisms operate. These discussions naturally bring along an array of terminology to refer to these systems, such as intertwined languages (Bakker 1994), converted languages (Bakker 2003), split languages (Myers-Scotton 2003), as well as a debate on their proper classification (Bakker 1994; Thomason 1995). Matras (2000 et seq) defines a prototype of mixed languages, in which one language effectively provides what he refers to as ‘predicate-anchoring’ and thus constitutes the INFL(ection)-language; i.e., the language that provides verbal inflection; while the other provides ‘most unbound, potentially autonomous content words, especially nouns’ (Matras 2003, p. 155). Concomitantly nominal morphology need not pattern with that of the INFL-language as well. The choice of the INFL-language “reflects the overall social and communicative orientation of the speaker” (Matras 2009, p. 305). In L1-shift cases, the INFL-language is the dominant language of the community, as is by and large the case for Ma’a, Angloromani or Light Warlpiri. In situations where speakers have partial resistance to language shift, the INFL-language is set as the ancestral language, yet selective copying of elements from the new language occurs (acculturation type, Matras 2009, p. 305), as in Media Lengua or Michif.

Matras’ model attributes a strong role to speakers as sentient beings mixing languages following deliberate community decisions. Decisions often associated with such mixture include the decision for new identity-building or the aim of maintaining secrecy, especially among various itinerant populations (Bakker 1994; but see also Bakker 1998). One of the lesser known mixed languages that are thought to have emerged in such a context is Lomavren, the mixed language of the Lom or Bosha/Posha, who dwelled in Armenia, Georgia and the northeastern part of Asia Minor in the beginning of the 20th century (Sargisyan 1864, p. 82; cited in Patkanoff 1908, p. 240; Finck 1907; Voskanian 2011). Lomavren, which was documented mostly in the early 20th century, is a mix of Armenian grammar and a significant amount of original Indo-Aryan vocabulary along with the retention of some Indo-Aryan functional morphemes (Voskanian 2011). Patkanoff (1908, p. 242) defines the then-status of Lomavren well: “The language of the Loms, which they are already forgetting, is exposed to such changes under the influence of Armenian, that, if I am not mistaken, most of their words take the grammatical forms of the Armenian language, only the roots of the words remaining”.

The commonly-held view on the emergence of Lomavren is built on the idea of language shift from some Central Indo-Aryan variet(ies) to Armenian and fixating the INFL-language as Armenian while retaining some inherited lexicon for establishing secrecy and an in-group identity (e.g., Patkanoff 1908, p. 244). This is definitely concomitant to changes in lifestyle and religion, for the details of which see Patkanoff (1908). At this point, we need to clarify that not every Posha or Lom community had the same linguistic (as well as social or religious) background in this vast region. Patkanoff (1908) illustrates this linguistic situation by stating that those in Artvin were ‘known to speak Turkish’ (p. 237) and those in Tokat ‘ha[d] a mixture of their own words with Turkish and Armenian’ (p. 243). It has been over a century since the last proper linguistic data were gathered from the Poshas in Asia Minor. This is no little time for a language to change, and focus on the linguistic practices of these groups is appropriate and timely. Following Üzüm’s (2012) pioneering work, our main objective with this study is to resume studies on the language(s) of the Poshas in Asia Minor. We initiate this by a close-investigation of the language of one particular group, the Posha of Çankırı in central Turkey. A crucial point to make clear at the outset is that this community’s members do not know the words Lom or Lomavren, and they refer to themselves and their language as Posha. We hereafter refer to their language as PoC, standing for Posha of Çankırı.

We provide an initial documentation of PoC, putting particular emphasis on points where it diverges from its allegedly parental language Lomavren, which was recorded in the early 20th century and a form of which is still spoken in Armenia today. This sketch grammar shows that PoC is a mixed language, just like Lomavren. In PoC’s case, however, the mixing is between Lomavren, whose grammar is largely from Armenian, and Turkish. PoC now exists symbiotically (à la Smith 1995) with Turkish, one of the source languages. While the verbal roots and verbal morphology are still recognizable as Armenian, in the nominal morphology and in the non-verbal vocabulary significant mixing occurs between Turkish and Lomavren. We further show that not everything can be attributed to one or the other source language directly. This suggests that PoC is an autonomous, self-contained and functional system (à la Bakker 2003), potentially independent of the rules of its respective source languages and subject to every diachronic change mechanism that a non-mixed language is subject to. We then turn our attention to the bilingual processes that contributed to the emergence of PoC. Here we show that a major contributor to the emergence of PoC is the fact that the speakers of its parent language shifted from Armenian to Turkish as the ‘new’ symbiotic language. The INFL-language of Posha, which is deemed the code that marks in-group identity and functions as a carrier of secrecy, is fixed as Lomavren, which, however, underwent minor changes; for instance, Indo-Aryan vocabulary was reduced even further. As the speakers ceased to be bilingual in Armenian, which once continuously fed into the grammar of Lomavren, and as societal bilingualism in Lomavren and Turkish increased, massive borrowing/copying (initiated as code-switching) from Turkish, the new symbiotic language, was inevitable. As a result, PoC seems to apparently combine Armenian and Turkish elements. Note, however, that the former are present only through Lomavren, the retained INFL-language.

The study is based on linguistic and ethnographic data collected in multiple fieldwork trips to Çankırı province (central Turkey) between 2010 and 2015. Linguistic data were mainly extracted from a corpus of naturalistic speech data, and were further enriched with information provided in questionnaires. Additional surveys and longitudinal observations within the community constitute sources to estimate speakers’ attitude towards PoC and its social functions with respect to the majority language, i.e., Turkish.

PoC has no standard orthography, nor has its phonology been investigated so that a principled transcription convention could be created. We, therefore, transcribe our linguistic examples only broadly, without necessarily complicating our examples. Most of the graphemes we chose for transcription are self-explanatory; we further use <ɨ> to render [ɯ], <č> for [tʃ] and <c> for [dʒ]. Vowel harmony across suffix boundaries exists but it is, at least for the time being, unpredictable and unsystematic. We nevertheless show harmonic suffixes with a capital letter. This denotes an archiphoneme whose surface value depends probably on the preceding vowel. It is only through a thorough study that we can see how vowel harmony works in PoC.

The paper is organized as follows: Section 2 provides information on the Posha community and Section 3 focuses on their linguistic practices. In Section 4, we situate PoC in its historical context to distinguish it from Lomavren. Section 5 provides the sketch grammar and Section 6 focuses on the expression of possession and nominal compound formation, two domains where mixing certainly occurs but where it is hard to define the sources. In Section 7, we elaborate on our opinion on the mechanisms behind its emergence. Section 8 concludes.

2. The Poshas of Çankırı Today

The linguistic community from whom the linguistic data were collected for this study identify themselves uniformly as Poshas. Communities that adopt the same ethnonym, i.e., ‘Posha’, are found across a large part of Asia Minor—especially in the northeast, mainly in the provinces of Artvin, Kars, Erzurum, Tokat and Erzincan, Gümüşhane (Bozkurt 2004; Üzüm 2012; Anumyan 2017). While many of these communities, including the majority of the Poshas of Çankırı we have interacted with, have clearly adopted a sedentary lifestyle (see also Arpacı 2019), there are still a few micro communities that maintain a peripatetic lifestyle, leaving their base settlements for neighboring cities in the summer months as itinerant traders. A few families among the Poshas of Çankırı too follow such a routine. That this commercial semi-nomadic lifestyle was probably once a reality for all Poshas of Çankırı can be inferred by the fact that many individuals, including those who are settled, refer to their ancestors as göçebes, which means ‘nomads’ in Turkish:

We settled here and we condemned the language of our own nomads [i.e., our ancestors, the authors].

(field notes, 28 October 2012)

Poshas are referred by the out-group as ‘gypsies’ (Turkish ‘çingene/cingan’), or are known with metonyms for the term ‘gypsy’, such as ‘sieve maker’ (Turkish ‘elekçi’) or ‘peddler’ (Turkish ‘bohçacı’) (Üzüm 2012; Demir and Üzüm 2014, 2018; Öncül 2019 among many others), which are terms that they think refer to their origin or occupation. These identifications are not immediately accepted by the group members and are found offensive, which is illustrated below with a quote from one informant:

We have no affinity whatsoever to the Gypsies. They [non-Poshas, the authors] are jealous of us, hence they slander us. I don’t like this at all. They bring along a five-year-old here and tell him that it is the Gypsy market. They ostracize us but they also get what they deserve. When their kids join the army, they are asked where they are from. When they say they are from Çankırı, they are immediately labeled as sieve-makers.

(Çetin 2014, p. 92, our translation)

Even as early as 1864, Sargisyan (1864) was hesitant to identify the Poshas of Asia Minor with Gypsies (Patkanoff 1908, p. 324).

The community adds up to roughly 2000 members. While sieve-making was the major occupation of the community until recently, only a few families depend on this today. Some families travel over large distances to sell almost anything they can, including fabric, carpets and rugs. A significant number of families nowadays own shops, mainly hardware or grocery stores.

The Poshas of Çankırı have largely remained under the radar in previous ethnographic studies on Posha groups in the Anatolian peninsula (e.g., Benninghaus 1991). We believe that one major factor contributing to this is the fact that, apart from their being a small community, they do not conform to the out-group’s typical characterization of them. They are not involved in the entertainment sector, something typically attributed to Poshas by the out-group. While this may indeed be the case for some Posha groups (Patkanoff 1908, pp. 237, 239), Çankırı Poshas do not engage in such practices as playing music and dancing in public, nor have they been involved in other entertainment sectors such as fortune telling or keeping dancing bears.

We do not dance or play music [for a living, the authors]. Our people live on selling goods in the villages. Our people are very keen on their dignity.

(Çetin 2014, p. 91, our translation)

Due to the fear of constant stigmatism from the out-group, the Poshas of Çankırı are often rather uneasy about revealing their in-group identities, something that also renders linguistic fieldwork difficult. When they are asked by the outsiders who they are, an oft-given answer is as follows:

We are from Çankırı. We do not have another language or religion. We all speak Turkish. We are Turks and Muslims.

(field notes, 2 December 2010)

Looking on the bright side, however, it is this stigmatism by the out-group that has enabled Poshas to conserve and transfer their cultural capital. There is a strong sense of in-group identity, which is also fostered by their religious identity: the Poshas of Çankırı are mostly adherents of Shiism, while the majority in Çankırı are identified as Sunni Muslims. Endogamy is the norm (also in the past, see Patkanoff 1908). Furthermore, children very often travel with their fathers to sell products, and therefore continue their parents’ profession. A corollary of this is the fact that the number of people undergoing formal education beyond the government’s requirement is rather rare. While all of these create a sense of in-group identity, we should, however, note that this in-group awareness should not be characterized as favoritism. There is no doubt that some resentment exists within the community towards Tartis or Poos, i.e., the Sunni Turkish people of Çankırı as they are called in the language of the Poshas; however, this is never formulated as hatred or disdain.

In the rest of the paper, we will simply use Poshas to refer to the focal group in Çankırı. Needless to say, the results of this study cannot be extended to other Posha communities across the country.

3. The Current Linguistic Situation of the Poshas

The Posha community can be characterized by and large as a bilingual one in Turkish and PoC. Apart from its use as an ethnonym, Posha is also the word they use to refer to their ancestral/home language, which we continue to refer to as PoC. They also have active command of Turkish, or Pooca (Poo-ish), as they call it in PoC (Üzüm and Demir 2017). The following quote by one informant summarizes this situation:

If you say twenty things in Posha, you say twenty other things in Pooca [i.e., Turkish, the authors]. You can’t always talk like this [=as a monolingual, the authors]! We could spend [the day talking in Posha, the authors], it is easy. We shout from here in the morning [to the neighbors, the authors]. We don’t shout in Turkish! We say, ‘Çile [personal name, the authors], Çile! Harɨt duni ye mi?’ (Posha: ‘Çile, Çile, is your father at home?), ‘Čile marɨt duni ye mi? (Posha: Çile, is your mother at home?), ‘Čile, hos ogo, hos ogo (Posha: ‘Çile, come here, come here’) […] One speaks their own language.

(fieldnotes, 9 March 2013)

It is however, only the elderly who often rely on PoC in intra-community interactions, whereas those in the younger generations, notably the ones below the age of 15, seem to only have some passive knowledge of it. For the elderly, as well, it is a home-language. The dramatic drop in the number of active speakers across generations stems from the continuous reduction in the functional domains of PoC by Turkish. PoC, however, still serves as a secret language in the presence of Poos:

May God protect us from bad days; if we have a secret to tell, we do it in Posha. This language protects us.

(fieldnotes, 31 January 2013)

As is common with the other minority languages in Turkey that are not enlisted, there is no external support system to increase the vitality of PoC. More dramatically, some Poshas are either indifferent to or in support of language loss, favoring linguistic assimilation to Turkish. PoC is confined today to mundane interactions within the group, which has also created a belief in the ‘defectiveness’ of their language: PoC, according to the community, has a limited lexical repertoire, and discussion of some topics is therefore impossible in the language. Despite this, the general picture is that PoC is largely regarded to have a function of uniting those who speak and understand it in a close-knit social group with its own characteristics.

Given the fact that PoC

(1)	a.	is not fully transmitted across generations,
	b.	is used in no other domains than mundane interaction or as a secret language,
	c.	is uniformly considered insufficient to meet the needs of other domains,
	d.	is not supported by the government, NGOs or other in-group bodies, and
	e.	has not been extensively documented, save the attempt by Üzüm (2012),

It should be classified as a severely endangered language with highly limited domains of use, according to the criteria determined by the UNESCO (Brenzinger et al. 2003).

4. The Posha and PoC in Its Historical Setting

The term ‘Posha’ has been used for a long time to refer to a group of people who have inhabited large parts of Asia Minor, especially the historical Armenian highlands. These people have been referred to as Bosha in Armenian but the historical autonym is known to be ‘Lom’. It is speculated that the Lom people, who are also called Armenian Gypsies in the western literature (Meillet 1921; Voskanian 2002; Scala 2014), migrated from northern India into eastern Asia Minor around the 11th century (Marsh 2008; Anumyan 2017; see Matras 2004, p. 16, for previous possible migration waves), which was possibly followed by other waves of migration. Clearly, Lom is an etymological cognate to Rom and Dom, words that designate other Indo-Aryan people of the Near East and Europe (Matras 2004). For Matras, the different but cognate names signal the existence of “repeated westward migratory ventures by individual groups seeking employment opportunities in specialized trades” (Matras 2004, p. 16).

According to some authors, the affinity between the Dom, Rom and Lom people also means linguistic affinity between Romani, Domari, and Lomavren, i.e., the language of the Loms. That all three can be attributed to a single language was claimed as early as in 1920s by Sampson (1927). The affinity is also judged plausible by Matras (2004), who claims, following Turner (1926), that all three might have originated in the Central group of Indo-Aryan languages (Matras 2004, p. 46; Voskanian 2011, p. 811). Lomavren, which is also known as Armenian Bosha, or Boshayeren to Armenians (Voskanian 2011), is thus rendered, at least originally, an Indo-Aryan language. When we look at its status since the early 20th century, Lomavren has been defined as a mixed language, consisting of inherited Indo-Aryan vocabulary inserted into the Armenian grammar frame (for earlier characterizations, see Paspati 1870; Finck 1907; Patkanoff 1908, for modern analyses see Voskanian 2002, 2011; Matras 2004; Scala 2014).

Leaving aside Sargisyan’s list of eighty-nine words (Sargisyan 1864), Finck (1907) presents the oldest available data on Lomavren. Finck futher claims that a massive lexical stock that is clearly of Indian origin is visible and some vestigial structures of Indo-Aryan are not completely lost (Finck 1907, p. 49). Patkanoff (1908) notes that Lomavren must be an Indo-Aryan language, based on its vocabulary (Patkanoff 1908, pp. 234–37, 245). Contrary to Patkanoff’s view, Andrews (1989, p. 140) thinks that gypsies living in eastern Asia Minor speak a dialect of Armenian. The disagreement about the classification of Lomavren is no more surprising than the difficulty in classifying other well-known bilingual mixed languages. In more recent literature, Lomavren is defined as a mixed language formed by Armenian grammatical structures and words of Indo-Aryan origin (Melikian (2002, p. 188). In a similar vein, Seropyan (2000, p. 23) explains the connection between Armenian and Lomavren as follows: “They usually use Armenian while speaking, but when they need to talk secretly among themselves, they use the Posha language. However, this language of Indo-European origin, which was getting poorer gradually, had to use Armenian in grammar and verb conjugation”. Scala (2014) compares Lomavren to Para-Romani varieties, and claims that the traditional Indo-Aryan is the lexifier code; however, unlike in Para-Romani varieties the traditional lexicon is retained massively and not through a process of selection.

Lomavren (Bosha) is claimed to be spoken in Armenia, Georgia and Asia Minor (modern-day Turkey) mostly as a secret language today (Voskanian 2002, 2011). Poshas we contacted today, however, state that they had never heard of the words Lom or Lomavren from their ancestors. They uniformly refer to themselves and their language as Posha. When they are asked if they are related to the Lom communities living in Armenia, they say that they do not know them. At this point, we choose to stay entirely agnostic about the ethnic origin of Poshas given their agnosticism about the term Lom. We are, however, certain that PoC and Lomavren (Bosha) have a strong historical affinity, so much so that we take PoC as a newly emerged variety of Lomavren in Çankırı.1 Apart from the obvious similarity between the names Posha and Bosha, this conclusion is based on the plethora of lexical items common to both Lomavren (Bosha) and PoC, the obvious dominance of Armenian in the grammar of both languages and the retention of Indo-Aryan and Iranian lexica. The difference lies in the fact that PoC has further undergone massive Turkish influence, which almost entirely replaced the Indo-Aryan lexicon and also had a substantial effect on the grammar. These are obviously missing in Lomavren as it has been reported in the literature. We therefore conclude that PoC has a strong affinity to Lomavren. It is a mixed language, and the mixing has happened between Turkish and Lomavren. In what follows we offer a brief introduction to the grammar of PoC, providing an overview of its lexicon, and a few notes on its nominal and verbal morphosyntax. The information on its morphosyntax is rather brief at this point; however, it serves well to support our characterization of PoC as a Lomavren-Turkish mixed language.

5. Grammar Overview of PoC

5.1. Lexicon

It has been stated in the literature that Lomavren’s lexicon is composed mostly of Indo-Aryan words (Finck 1907; Voskanian 2002; Scala 2014). The rest of the lexicon is taken from Iranian languages and Armenian. Some of the Iranian words might have entered the language via Armenian (Voskanian 2002). While these studies do not necessarily make explicit whether the above partitioning is valid for all lexical categories or not, one thing that is clear is that Indo-Aryan vocabulary is still salient.2 This picture is rather different in PoC. The classes of nouns, adjectives and adverbs are almost wholly filled by words of Turkish and Armenian origin, while Indo-Aryan words are simply vestigial. Impressionistically, words of Armenian origin dominate in all major lexical categories and in every semantic domain. Some of the commonly attested words from Armenian are dun ‘house’ (<W(estern) A(rmenian) dun), gadu ‘cat’ (<WA gadu), campa ‘road’ (<WA dʒampʰa), va ‘tomorrow’ (<WA vaʁə), bax ‘cold’ (<WA baχ) and hos ‘here’ (<WA hos, see DDAL 2001, p. 301 for specific dialects).3 Among the few Indo-Aryan words that we could detect that are common to both Lomavren and PoC are kɨl ‘oil’ (cf. Lom. kyl ‘butter’ (Patkanoff 1908, p. 254), cf. khil (European Romani)) and innav ‘boy’ (cf. Lom. hinnav ‘a young man’ (Patkanoff 1908, p. 251)). While numerous Turkish words do appear in all lexical categories, they are commonly attested in the class of nouns and conjunctions. Some words of Turkish origin are torun ‘grandchild’ (<T(urkish) torun ‘grandchild’), erkenden ‘early’ (<T erkenden), sɨğɨn ‘horse’ (<T sığın ‘deer’), kayfe ‘coffee’ (< T kahve), Cuma ‘Friday’ (<T Cuma), hem… hem… ‘both … and …’ (<T hem… hem…), ha…ha… ‘whether…or…’ (<T ha…ha…). Attributing therefore a direct diachronic link between Lomavren and PoC, we conclude that Turkish or Armenian vocabulary has almost entirely replaced Indo-Aryan. Gughvi ‘cold’, for instance, which is reported as a Lomavren word by Patkanoff (1908, p. 255), is no longer recognized by PoC speakers; informants uniformly use the word bax to express the same meaning (<WA baχ ‘cold’). Among many other such lately-appeared Armenian or Turkish words are čur (<WA tʰʃuɾ), which replaced pani ‘water’ (Patkanoff 1908, p. 253) and ineg ‘cow’ (<T inek), which replaced mozlax (Patkanoff 1908, p. 252). Furthermore, almost all the concepts expressed with inherited Indo-Aryan words in PoC can also be expressed with Armenian words. This is most notable in the first-level kinship terms. Examples are given in (2).

(2)	a.	son: dɨğa (<WA dəʁa) & innav (cf. Lomavren (h)innav, Patkanoff (1908)),
	b.	father: har (<WA haɾ (see DDAL 2001, p. 254 and Adjarian 1909, p. 45) & dad
		(cf. European Romani dad, Hindi tat),
	c.	daughter/little girl: axčik (<WA aχtʰʃig, see Martirosyan 2009, pp. 35–36 for various dialectal forms) & coki (cf. goqi ‘a young girl, child, a sister’, Patkanoff 1908, p. 251, cf, čhoři in various Romani dialects, ROMLEX 2022).

While PoC still retains a small Indo-Aryan core lexicon in the domain of substantives (in which words of Turkish origin also abound), verbal predicates are exclusively of Armenian origin. Some examples of verbal predicates are bargel ‘to lie down’ (<WA baɾgel), kal ‘to come’ (<WA kʰal), bahel ‘to hide’ (<WA bahel), pakčel ‘escape’ (<WA pʰaχtʰʃil), anal ‘make’ (<WA ənel,; WA.dial (Erzurum) enel; see Greppin and Khachaturian 1986, p. 91; E(astern) A(rmenian) anel).4

5.2. Notes on Nominal Inflection

Unlike numerous other Romani varieties, for which see Elšík (2000) and Matras (2004), and similar to Lomavren, there are no nominal inflectional classes that are based on grammatical gender in PoC. Nouns do not encode morphological gender, nor do they control gender agreement. As a corollary, modifiers of the noun appear in invariant forms, and they are in prenominal position. Like in Lomavren (cf. Voskanian 2002, p. 171; 2011), vestigial forms of the historical Indo-Aryan masculine or feminine endings appear in a few inherited Indo-Aryan words as masculine -av and feminine -i. Compare these to -o and -i in Romani. Such examples include pusav/pɨsav ‘man, gičav ‘gun’, (h)innav ‘son’, tatav ‘rain’ or coki ‘daughter/little girl’. These, however, do not justify proposing different inflectional classes; synchronically they are parts of the stems. While singular number in nouns has no overt morphology, plural number is expressed by –diy and –er suffixes (Üzüm and Demir forthcoming): -er is attached to monosyllabic roots (3) whereas –diy appears on polysyllabic ones (4):

(3)	a.	hav-er-i	b.	maz-er-i	c.	šun-er-i
		chicken-pl-def		hair-pl-def		dog-pl-def
		‘chickens’		‘hairs’		‘dogs’
(4)	a.	erimart-diy-i	b.	okkaman-diy-i
		person-plA-def		shoe-pl-def
		‘people’		‘shoes’
	c.	tarihi eser-diy-i
		historical artifact-pl-def
		‘historical artifacts’

Both morphemes are of Armenian origin; -er is from –(n)er, which is common to WA and EA; for an extensive discussion of the origin of the -diy form see Scala (2014, pp. 242–44).5 In the early 20th century, Finck (1907) noted the existence of similar plural suffixes in Lomavren (Scala 2014)—albeit with minor phonological differences. Scala (2014) clarifies that in Finck’s database words that do not end in -av receive -(v)avtik^h, while for those that end in -av, both -ner and -tikʰ suffixes are possible as plural markers. We do not know if this means that they were in free-variation. Voskanian (2011, p. 812) reports for modern-day Lomavren (in Armenia) only the existence of the -tikʰ/vatikʰ form. In these respects, PoC plural marking differs from that of Lomavren, both the one recorded by Finck and the one spoken currently in Armenia.

Morphological case in PoC is expressed for the genitive, dative and ablative (see also Üzüm and Demir forthcoming). We have not been able to verify the existence of the instrumental case, which is noted to exist in Lomavren, in PoC. The nominative and accusative have no overt exponents. The case system of PoC is given in

Table 1. PoC case system.

Case	Exponent
nom	-Ø
acc	-Ø
gen	-(y)In
dat	-(y)e
abl	-(y)en

for singular nouns:

The suffixes in Table 1 are illustrated in context in (5):

(5)	a.	Sinek-Ø	mud-av.	(nom)
		fly-nom	enter.pst.3sg
		‘(A) fly entered.’
	b.	Ses-Ø	m-aneg!	(acc)
		noise-acc	mneg-make.imp.2pl
		‘Don’t make noise!’
	c.	gadun-un	por-ɨ…	(gen)
		cat-gen	stomach-def
		(the) cat’s stomach…
	d.	Coki-ye	kafe per!	(dat)
		girl-dat	coffee bring.imp.2sg
		‘Bring coffee to the girl!’
	e.	Os-en	gɨ-vahɨnmor.	(abl)
		snake-abl	ipfv-fear.1sg
		‘I am afraid of snakes’.

No overt exponents are present for the nominative and the accusative; nouns whose referents are generic/indefinite in subject and object positions appear in bare forms. This syncretism exists in Lomavren as well (Voskanian 2011). Subject/object nouns with definite referents are overtly marked with the definiteness marker -I or -n; the former appears on consonant-final words and the latter appears on vowel-final ones. The examples are provided below:

(6)	a.	[An	dun-i]	as dun-en	mez = e.	(subject)
		that	house-def	this house-abl	big = cop.3sg
		‘That house is bigger than this house.’

 

	b.	[Tur-ɨ]	pats!			(object)
		door-def	open.imp.2pl
		‘Open the door!’

The definite articles are present in Lomavren as well (Voskanian 2002, p. 171) and are clearly of Armenian origin, in which they appear as -ə and -(i)n:

(7)	a.	[Ajs	kʰoɾkʰ-ə]	noɾ = e.		(subject)
		this	carpet-def	new = cop.3sg
		‘This carpet is new.’

 

	b.	[seʁan-in]	zargav.			(object)
		table-def	hit.pst.3sg
		‘(S)he hit the table’				(WA)

The distribution of the definite article and the factors conditioning its allomorphy are complex in (Western) Armenian and are closely linked to differential object marking (see for the details Khanjian 2013, p. 32ff; Donabedian-Demopoulos 2018). Whether or not there is isomorphism between PoC and Armenian in terms of the marking of definiteness remains to be seen. Lomavren (Scala 2014, p. 239, Voskanian 2011, p. 812) and Armenian (Donabedian-Demopoulos 2018, p. 17) are both reported to possess a singular postnominal indefinite marker, essentially the same morpheme, as in mə:dun = mə ‘a house’ (WA) and bar-mə ‘a door’ (Lomavren). This does not exist in PoC, where indefiniteness is expressed by the prenominal numeral mek ‘one’.

As for the origin of the cases in PoC, we need to compare them with their Turkish and Armenian counterparts as well as those in Lomavren as it is spoken in Armenia and Georgia (Voskanian 2011).

It has been stated in the literature that Lomavren’s morphology is from (some dialects of) WA (see Voskanian 2011, p. 811, for a recent claim on this and see Scala 2014 for an assessment), and this naturally includes nominal morphology, as

Table 2. Case inventories of PoC, WA, EA and Turkish.

Case	PoC	Lomavren	WA6	EA7	Turkish
nom	− Ø	− Ø	− Ø	− Ø	− Ø
acc	− Ø	− Ø	− Ø	− Ø	− Ø/− (y)I8
gen	− (y)In	− i	− i	− i	(n)In
dat	− (y)e	− i	− i	− i	− (y)A
abl	− (y)en	− e	− e	− ic’	− DAn
inst	−	− ov	− ov	− ov	− (y)lA
loc	−	−	−	− um	− DA

clearly illustrates. The case system of PoC is ‘simpler’ with respect to that of Lomavren, as the instrumental case seems to be missing. Ablative is the reanalyzed form of the Armenian ablative -e and the Armenian definite article -n (8). In particular, (9b) shows that –(y)en is a monomorphemic suffix, encoding no definiteness.

The nominative–accusative syncretism in PoC is a successor of the syncretism in Lomavren and its origin is clearly Armenian. The syncretism between the dative and the genitive in Lomavren and in Armenian, which also indicates that the respective case forms of Lomavren are from Armenian, is missing in PoC: the dative and the genitive have distinct exponents. While the formal resemblance between the Turkish and the PoC forms of the genitive case is clear, in Section 6 we will further discuss this and conclude that one cannot simply trace this to Turkish or Armenian. The dative case is provided by Turkish, although the suffix does not seem to be vowel-harmonic in PoC. As such, PoC case morphology today is a fusion of Turkish and Lomavren—and therefore indirectly of Armenian.

(8)	[[[kʰaɾ] -e] -n]
	stone -ABL -DEF
	‘from the stone’	(WA, cf. kʰaɾ-e [stone-abl] ‘of/from stone(s)’)

 

(9)	a.	os-en	b.	kahk-en
		snake-abl		where-abl
		‘of/from (the) snake(s)		‘from where? (PoC)

5.3. Notes on the Origin of the Verbal Inflection and Negation

PoC is as a language with a nominative–accusative syntactic alignment system, based on verbal agreement patterns and word order. The basic word order is SOV, with subject (Sole/Agent) –verb agreement only. Tense/aspect/modality inflection and person agreement markers, as well as the copular form, are exclusively of Armenian origin. In (10) below, Armenian counterparts of the predicates are given in parentheses.

(10)	a.	Emmen	Cuma	torun-diy-is	gu-ka.
		every	Friday	grandchild-pl-poss.1sg	ipfv-come.prs.3pl
		‘My grandchildren come (to me) every Friday.’			(cf. WA gu kʰa)

 

	b.	bargel-i		=yem…
		lie-ptcp.fut		=cop.1sg
		‘I will lie down…’
			(cf. WA.dial. baɾgelu jem (Erzurum, Greppin and Khachaturian 1986, p. 91)
	c.	Gičav-i		bahe!
		gun-acc		hide.imp.2sg
		‘Hide the gun!’			(cf. WA. dial. bahe (Erzurum))

Similar to the case in all current Romani varieties, clausal negation is sensitive to mood in PoC. The standard negation marker in sentences in the indicative mood (ineg) is the verbal prefix č(I)- (11a) whereas negative sentences in the non-indicative mood, i.e., prohibitives (mneg), are marked with the prefix m(I)- (11b):

(11)	a.	Öküz-diy-i	či-ga.
		ox-pl-def	ineg-exist.prs.3
		’The oxen are not here.’

 

	b.	Ses	m-aneg!
		noise	mneg-make.imp.2pl
		‘Don’t make noise’

The indicative negation marker is certainly from Armenian (<WA tʰʃ(ə=)), a language that also makes a distinction between indicative and non-indicative negation, using the preverbal morpheme mi in the second case. Based on the indicative vs. non-indicative dichotomy and on the formal resemblance between the forms, one could also argue that the non-indicative negation marker in PoC is also of Armenian origin. At this point, however, we should note that the form of the negator in all current Romani varieties is sensitive to the mood of the verb. They minimally make a two-way distinction between indicative and non-indicative negation types (Elšík and Matras 2006, p. 158; Matras 2004, p. 189 a.o.). The latter is expressed by ma=, a form which is formally similar to the non-indicative negation marker in PoC. Given these facts, we consider it plausible that the two-way distinction in negation, as well as the overt exponent of negation with verbs in the non-indicative mood, are inherited from Indo-Aryan. Circumstantial evidence for this comes from Lomavren, which we take as a cognate/precursor of PoC. In Lomavren as well, clausal negators are sensitive to the mood in the clause, exactly as in Armenian, Romani and PoC. However, unlike in PoC, both negators are inherited: na = in indicative sentences and ma = in prohibitives (Scala 2014, p. 242; Voskanian 2011, p. 816). We show the negation markers in the relevant languages in

Table 3. Negators in Armenian (WA/EA), PoC and Romani.

NEG	PoC	Lomavren	Armenian	Romani
indicative	č(I)=	na=	tʰ∫(ə)=	na=
prohibitive	m(I)=	ma=	mi=	ma=

.

While the two-way distinction of negation in Lomavren and PoC is an inherited feature, its maintenance is certainly fed into by Armenian. On the other hand, if we are to take PoC as a variant that emerged from Lomavren, we should also posit that at least the indicative negator is replaced by the Armenian one at a later stage.

The above few notes on the origin of inflection in the verbal domain in PoC imply that the Turkish influence never made its way into this domain—hence, PoC verbal morphosyntax is Armenian with a possible though faint layer of Indo-Aryan. What little Turkish verbal/predicate morphology we observe in the database always appears on Turkish verbal roots:9

(12)	a.	Pɨsav	gel-iyor-Ø.
		man	come-prog-3sg
		‘There comes a man’		<T. geliyor

 

	b.	Yara-s	yürek-te-dür. heart-loc-mod
		wound-poss.1sg
		‘My wound is in (my) heart.’		<T. yürektedir

We take such cases as (12), where the entire predicates are in Turkish, to be instances of code-switching in which the Matrix Language (cf. Myers-Scotton 1993 et seq) is defined by the predicate. Phrases with full inflection in the other language logically appear as Embedded Language Islands within the Matrix Language, as is the case in (13) and (14). In (13), the subject NP is entirely from Turkish, along with all the inflectional suffixes, whereas the inflected verb is native. In (14) two sentences (inflected verbs) are combined by the PoC clause connector ha…ha…, which is ultimately Turkish as well. Crucially, the first one is Turkish and the second one is PoC.10 The examples in (12)–(14) are also an indication that code-switching is quite common within the community.

(13)	ırbığ-ın	arka-sın-ı	kise!
	water.pitcher-gen	behind-poss.3sg-acc	clean.imp.2sg
	‘Clean the back of the water pitcher!’		<T. ibriğin arkasını

 

(14)	Ha yi-di-Ø	ha č-udi.
	Either eat-pst-3sg	or ineg-eat.subj.3sg
	(‘It does not matter if) (s)he eats or not.’		<T. yedi
			<WA tʰ∫-ude

In the last three subsections, we have shown that what little lexical or morphosyntactic Indo-Aryan elements were/are present in Lomavren have been reduced even further in PoC. The latter’s lexicon is almost entirely composed of words of Lomavren and Turkish origin, but, crucially, there is a dichotomy in the system: Turkish verb roots do not exist in PoC, despite the by-now long-term bilingualism of the speaker community. Likewise, verbal morphosyntax, at least the overt forms of the exponents, are entirely from Lomavren, which apparently have replaced the inherited, Indo-Aryan categories (e.g., the exponents of negation) as well. Although we can trace the origin of the categories back to Armenian, only a thorough investigation will reveal whether the functions of verbal morphemes are isomorphic between PoC and Armenian. While Turkish never penetrated the verbal domain, its influence is present in the nominal morphosyntax of PoC; see for instance the overt exponent of the dative case. Innovations (as loss of categories) are also observed in PoC, such as the loss of the instrumental case. These facts indicate that, while PoC and Lomavren are probably historically related, PoC is a linguistic system on its own today, and is one that can most prominently be characterized as a mixed language.

In the next section, we will return to the nominal morphology again, focusing this time on the expression of possession and, related to this, the formation of nominal compounds. The reason why we look at these domains in a separate section is that they are particularly interesting ones in revealing PoC’s mixture of Turkish and Armenian.

6. Expression of Possession in PoC

PoC typically marks possession on the possessed noun by suffixes that agree in person and number with the possessor. This is a trait shared by both Armenian (WA) and Turkish. The set of possessive suffixes in PoC are immediately recognizable as Armenian. Compare the PoC, WA and Turkish possessive suffixes in the singular in

Table 4. Possessive suffixes in PoC, WA and Turkish.

	PoC		WA		Turkish		Translation
1	-(I)s	adun-us	-(ə)s	anun-əs	-(I)m	ad-ım	‘my name’
		adun-ɨs
2	-(I)t	adun-ɨt	-(ə)t	anun-ətʰ	-(I)n	ad-ın	‘your name’
		adun-it
3	-I	adun-ɨ	-ə	anun-ə	-(s)I	ad-ı	‘his/her/its name’
		adun-i
	-n	diga-n	-n	dəʁa-n		oğl-u	‘his/her/its son’

:

There is allomorphy of the third person possessive (-I ~ -n) in PoC, similar to the allomorphy in WA. The latter form appears on V-final roots, whereas the -I form appears exclusively on C-final ones.

The possessive suffixes follow the plural:

(15)	a.	argi-diy-it	b.	torun-diy-is
		tooh-pl-poss.2sg		grandchild-pl-poss.1sg
		‘your teeth’		‘my grandchildren’

As is the case in WA, the third-person possessive morphemes are syncretic with the definiteness morphemes in both PoC and Lomavren (for the expression of definiteness in PoC, see Section 5.2). Apparent ambiguities are resolved in PoC contextually and on the semantic properties of the noun (see Donabedian-Demopoulos 2018, p. 18). (16a); for instance, there may be ambiguity between a reading in which two definite entities are present in the utterance context and a reading in which a third person possesses the two entities. Notice that the verb gam (<WA gam) can be used both as a be-existential, and as a have-existential with possessive suffixes on the relevant noun (see (16b) for the latter):

(16)	a.	Hem axčig-i			ga	hem dɨğa-n	ga.
		both daughter-def/poss.3sg			exist.3sg	and son-def/poss.3sg	exist.3sg
		‘S/he has both a son and a daughter.’
		‘Both the son and the daughter are (here)’.
	b.	Mek	tene	zavağ-ɨs	ga.
		one	piece	son-poss.1sg	exist.3sg
		‘I have one son.’

Like in Turkish and in Armenian, the possessor, marked with genitive morphology, can occur with the possessed noun either for emphasis, or—for the third person—for identification. These possessors precede the possessed nouns:

(17)	a.	imin yara-s	b.	tuyɨn dun-ɨt
		my wound-poss.1sg		your house-poss.2sg
		‘my wound’		‘your house’

The singular pronominal possessors marked in the genitive (hereafter possessive determiners) are given in

Table 5. Personal pronouns and possessive determiners in PoC, WA and Turkish.

	PoC		WA		Turkish
	Pronoun	Determiner	Pronoun	Determiner	Pronoun	Detserminer
1	es	im-in	jes	im	ben	ben-im
2	tu	tu-yɨn	thu	khu	sen	sen-in
3	an	?	an	iɾ (/an-oɾ)	o	on-un

, along with the WA and Turkish counterparts. For each language we also provide personal pronouns in the nominative. (see Scala 2014, p. 244; Voskanian 2011, pp. 813–14 for an entirely different set of pronouns in Lomavren):

The following notes on Table 5 are in order: WA possessive determiners and personal pronouns in the nominative are suppletive forms, e.g., jes ‘I’ vs. im ‘my’, whereas the Turkish possessive determiners are paradigmatically built on the nominative (bare) forms by adding the genitive suffix –(n)In, which appears as the allomorph -im on the first-person pronoun ben. Personal pronouns in PoC are uniformly from WA. The possessive determiners have an agglutinative structure. The first-person determiner is formed on the monomorphemic possessive determiner of WA, im, while the second-person possessive determiner is built on the bare/nominative form of the personal pronoun (namely tu). We could not elicit the third person determiner; speakers chose to refer to the possessor by name adding the genitive suffix, e.g., Čile-yin [Čile-gen] ‘Čile’s’. In both cases, the same genitive suffix attaches to the root. The PoC possessive determiners bring to mind the possibility that both the strategy of forming them and the genitive morpheme are of Turkish origin. This may very well be the case, but we further note that that there is no isomorphism between the Turkish and PoC genitive morphemes: PoC does not show the allomorphy that Turkish does between first person and non-first persons, nor does the suffix undergo vowel harmony the way its Turkish counterpart does. Furthermore, the same genitive suffix appears on non-pronominal nouns in Turkish, in which case, if the word is a V-final one, a [n]-segment precedes the suffix, for instance, keçi-nin [goat-gen] ‘goat’s’. Recall from Section 5.2. that in the same environment in PoC, the genitive suffix appears as -yIn. We will elaborate on this below. To conclude here, even if one would claim that the Turkish genitive and the agglutinative morphology are the sources of the pattern in PoC, this would only be an instance of selective copying (in Johanson’s terms, Johanson 1992 et seq).

That copying from Turkish would be selective is also evident by the fact that the genitive in the third person in PoC appears uniformly as –(y)In on V-final roots, unlike the Turkish form -nIn in the same context. Compare the PoC genitive form with the Turkish one in (20):11

(18)	a.	keçi-nin	[Turkish]
	b.	keči-yin	[PoC]
		goat-gen
		‘(the) goat’s’

The lack of initial [n] in the genitive when it is added to V-final roots may be taken as further support for the claim that the genitive is not (entirely) from Turkish. At this point, however, we should also address the question, ‘Which Turkish dialect is the source variety for the formation of PoC?’ The question becomes particularly relevant when we consider that in certain Turkish varieties the genitive appears as –yIn when added to V-final nouns, hence with a glide instead of the nasal in initial position:

(19)	masa-yın
	table-gen
	‘(the) table’s’
		(Turkish dialect of Kars, Susuz (Buran 1996), cf. Standard Turkish masa-nın)

The glide-initial case form has not been reported in the literature for the Turkish spoken in Çankırı. On the other hand, it may have been the case that a Turkish variety other than that of Çankırı was the source for many features of PoC. Given that we do not know the migration history of the Posha people of Çankırı, we cannot completely rule out the possibility that the genitive has emerged from some local Turkish dialect.

On the other hand, an account that attributes the emergence of agglutinative forms purely on code-copying from Turkish may be errant, or at best incomplete. It is known that in colloquial WA the definite article frequently follows the genitive/dative given that possessors are mostly definite entities in discourse. In this case the definite article appears as -n:

(20)	a.	Aɾam-i-n	namag-ə
		Aram-gen-def	letter-def/3poss
		‘Aram’s letter’
	b.	hamalsaɾan-i-n	bamagan	ʃenkh-eɾ-ə
		university-gen-def	historical	building-pl-def/3poss
		‘the historical buildings of the university’
		(adapted from Donabedian-Demopoulos 2018, pp. 15–16)

It is possible that the genitive suffix in PoC stems from ‘the genitive + definite article’ sequence of WA. With morphological boundary erosion, the definiteness and the definiteness suffixes have been reanalyzed as one single suffix, just like the ablative case suffix (cf. Section 5.2).

(21)	a.	WA: [[N -i] -n]
		-gen-def
	b.	PoC [N -in]
		-gen

Under this account, the rise of the agglutinative genitive forms is simply a simplification/optimization strategy where pronouns and nouns appear uniformly with separable affixes when the genitive is expressed on them.12

That we cannot precisely trace the PoC genitive to Turkish or to Armenian suggests that PoC is not a plain combination of Turkish and (via Lomavren) Armenian grammatical structures and elements. In other words, the whole does not directly follow from the sum of its parts. This indicates that, while some constructions or morphemes can be (surely or vaguely) traced back to Armenian or Turkish, PoC is still a complete system on its own, having its own historical trajectory and its speakers being active agents in shaping the language. To substantiate this claim further, let us look at nominal compound formation in PoC. The genitive suffix -(y)In is employed in all types of N+N compounds excluding co-compounds (subordinative, some attributive or quotative compounds). Such compounds are right-headed and the genitive attaches to the non-head constituent:

(22)	a.	keči-yin	gat-i
		goat-gen	milk-def
		‘the goat’s milk’
	b.	Salɨ-yɨn	or-ɨ
		Tuesday-gen	day-def
		‘(on) Tuesday’
	c.	Va-yɨn	sabah-ɨ
		tomorrow-gen	morning-def
		‘tomorrow morning’

All these examples are from the spoken corpus. We deliberately gloss the final morphemes in each example as ‘definite’ rather than possessive. The latter would definitely be illogical, at least for (22b) and possibly for (22c). In (22a), it necessarily brings definiteness to the entire concatenation. Under controlled elicitation, one further verifies this: compounds without definiteness at the end are also possible. We therefore conclude that the compound template in PoC is [N-GEN N].

The [N-GEN N] template for compound formation is not attested in Turkish. While there are a few details which need not concern us here, in Turkish nominal compounding relies on the use of a compound marker that has originated from the third person possessive and is attached to the right-hand constituent, which is the head. Compounds in (22a-b) are rendered in Turkish as follows:

(23)	a.	keči	süt-ü
		goat	milk-cm
		‘goat’s milk’
	b.	Salı	gün-ü
		Tuesday	day-cm
		‘(on) Tuesday’	[Turkish]

(22c) is treated as a plain attributive structure in Turkish and appears without the compound marker:

(24)	c.	yarın	sabah-(*ı)
		tomorrow	morning-(*cm)
		‘tomorrow morning’		[Turkish]

In Armenian, while an N-GEN N structure is available, it can express only subordinative compounds (and certain attributive compounds, for EA see Dum-Tragut 2009, pp. 83–84); thus, neither (22b) nor (22c) would appear in this structure. See the Armenian counterparts of (22) in (25):

(25)	a.	ajdz-i	gatʰ
		goat-gen	Milk
		‘goat’s milk’
	b.	jerekʰʃaptʰi	oɾ-ə
		Tuesday	day-def
		‘(on) Tuesday’
	b.	vaʁə	aɾavodjan
		tomorrow	morning
		‘tomorrow morning’		(WA)

Even if one would like to claim an Armenian origin for the template [N-GEN N], its productivity, unlike its Armenian counterpart, suggests that the copying is only selective. On the other hand, the template may even be a retention from an older stage of the language. In this scenario, the inherited [N-GEN N] template, in which the left-hand constituent functions morphologically as an attribute, is retained in PoC; however, as in the rest of the morphological domain, the compound marker is now of Armenian/Turkish origin. Compounding with the [N-GEN-N] structure is claimed to be a “[a] Common Romani structural resource” by Matras (2004, p. 77), viz. (26):

(26)	bakr-esk-o	Mas
	lam-gen-m	Meat
	‘lamb’s meat’	(Romani, Matras 2004, p. 76)

We then conclude that the source language for the compounding template cannot be straightforwardly identified. No matter what the origin, its productivity is clearly a PoC feature, which once again tells us that PoC cannot be reduced to a combination of linguistic features from two or more languages.

7. PoC as a Mixed Language

The above admittedly cursory notes on the inflection in PoC have established that PoC entertains an autonomous status as a system. This is further verified by the stance of its speakers, who distinguish it from any linguistic varieties they are aware of, and who view it as an established part of their identity. The data further revealed that PoC largely fits in the definition of mixed languages (Thomason and Kaufman 1988, p. 12; Matras and Bakker 2003 a.o.), i.e., languages in which lexical or grammatical material from more than one source language is combined in a systematic way, the choice being no longer subject to social or discourse facts (O’Shannessy 2021, p. 325). It is of no doubt that PoC combines grammatical material from Lomavren and Turkish, with possible minor retentions from Indo-Aryan. It largely retains the parental Lomavren (<Armenian) verbal morphology but has gone even further than Lomavren in incorporating Armenian elements into this domain. We have seen one example of this: the PoC negation morphemes are (possibly) both from Armenian, while Lomavren retains the Indo-Aryan ones. Similarly, although Lomavren was reported to include some Indo-Aryan verb roots (Voskanian 2002; Scala 2014), PoC has solely Armenian verbs. In nominal morphology too Lomavren is immediately recognizable; however, here we find both sui-generis innovations, e.g., plural marking acquiring sensitivity to animacy and the elimination of the instrumental case, and innovations that have been brought in by Turkish, e.g., the dative marker (and possibly the genitive too). The non-verbal lexicon is shared between Armenian and Turkish, native Indo-Aryan words simply being vestigial. All these, we suggest, render PoC a mixed Turkish-Armenian language at face value. While claiming this, we are also aware that Armenian in this very sentence refers to the Armenian stock of grammar of Lomavren.

The emergence of PoC brings to mind the other members of its kin, namely Lomavren and many Para-Romani varieties: that it largely involves L1–L2 shift, with an attempt by the speakers to retain a secret mode of speech (Bakker 1994; see Matras et al. 2007 for a good elaboration). We think, however, that the shift story only marks the beginning of the emergence of PoC. Unlike Para-Romani varieties or Lomavren, PoC seems to have gained a fossilized/autonomous status at a very early stage of this shift.

First, early differences in local Lomavren varieties were already evident a century ago. Consider, in this context, Patkanoff’s reports of two Lomavren varieties. While Patkanoff (1908, p. 243) reported that “instead of a language, [Poshas of Tokat, the authors] ha[d] a mixture of their own words with Turkish and Armenian”, those in Shirak spoke “(if one may so express it) a stepson of the Armenian language” (Patkanoff 1908, p. 244). It is therefore plausible that the early speakers of PoC/Lomavren in Çankırı probably shifted more to L2 Armenian, which means that it incorporated more Armenian grammatical elements than the Lomavren varieties reported in the literature. An example of this is the Armenian negation markers in PoC, which are absent in the literature on Lomavren, both the synchronic one and the diachronic one(s).

Apart from the degree of shift/ancestral language loss, the major force that distinguished PoC from the rest of the Lomavren varieties was the shift from one symbiotic language, Armenian, to another one, Turkish. The shift was inevitable due to obvious changes that happened inside the country with the establishment of the republic. The result that it bore is a Posha community whose members are all are active speakers of Turkish today; some, mostly the younger, are monolingual in Turkish and know only a limited number of vocabulary items in PoC. Crucially, however, rather than resulting in the death of their ancestral language, this shift instead created a bilingual society for which the function domains of the two languages were clearly demarcated. The Poshas, being already marginalized by the out-group, (still) possessed the need for a language that helped them maintain a close-knit community with its own set of values and a means to maintain secrecy to maximize benefit from any given situation (cf. Bakker 1994). The medium that provided these was established by retaining (at least a substantial portion of) the parental Lomavren grammar, to which the out-group members did not have access (see also Seropyan 2000). Notice that in this context PoC is different from many Para-Romani varieties. In Anglo-Romani, for instance, it is the grammar of L2 into which words from the ancestral language are inserted for similar purposes (cf. Matras 2009, pp. 295–96). That PoC has the main function of group formation and more importantly of keeping things secret is also a recurring topic that the informants continuously refer to (see Section 2 on this). The language’s function of keeping things secret naturally extends to keeping the language as a secret as well. For a very long time after the field work began, no one was willing to teach a word of PoC to the principal author. One question that she often received was as follows:

What are you going to do with our language? Did you come here to bring our houses down?

(field notes, 2 December 2010)

The attempt at retaining the ancestral language as a secret one did not yield a grammar that is identical to that of their ancestral language. Being bilingual, speakers would introduce Turkish elements when for some reason the ancestral ones (whether of Armenian or of Indo-Aryan origin) did not come to them. In other words, we claim that PoC, while started as a shift from L1 to L2, was in the end shaped by borrowing (through code-switching) from L2 to L1. That no Turkish effect is seen in the verbal morpho-syntax can be explained, in line with Matras (2009), if we take predication as the real symbol of the bilingual speakers’ context-bound choice of language. In the end, the outcome, i.e., the mixed language, functions as an emotive mode of its bilingual speakers (Matras et al. 2007), a mode of speech that signals intimate attitudes, knowledge and need, shared only by members of the community.

8. Conclusions

In this paper, we have provided a few notes on the grammar of a mixed language spoken currently in Çankırı, Turkey, based on data gathered through interaction with its native speakers. The data revealed that the source languages are Turkish and apparently Armenian. We say ‘apparently’ because mixing in fact has happened between Turkish and Lomavren, the latter of which is itself a mixed language between Armenian and some native Indo-Aryan varieties in the remote past. In Posha of Çankırı, the native Indo-Aryan lexicon was largely replaced by Armenian and Turkish words; more importantly, Turkish also made its way into the domain of nominal inflection. Crucially, however, the verbal domain remained intact, and in actual fact converged even more with Armenian than its parent language. We discussed that the emergence of PoC started with the usual mechanism of language shift (to Turkish), but at the same time speakers needed to retain a secret means of communication. The result is different from many Para-Romani varieties. In PoC, the inherited Lomavren (<Armenian) grammar was kept largely intact, especially in the verbal domain; however, Turkish elements, which are introduced via code-switching, became institutionalized parts of its grammar.