Basic vocabulary of closely related languages in contact: case study of Turkic languages on the Crimean Peninsula

Размещено на http://www.allbest.ru/

Размещено на http://www.allbest.ru/

School for advanced studies in the humanities

Basic vocabulary of closely related languages in contact: case study of Turkic languages on the Crimean Peninsula

Ilya Egorov

Moscow

Abstract

The present paper provides two case studies of the basic vocabulary of the Turkic languages spoken on the Crimea Peninsula. Its aim is to illuminate the issues that a historical linguist, and in particular a phylogeneticist, faces when analyzing the basic vocabulary of closely related languages in a situation of intensive contact. The first case study is dedicated to the onomasiological reconstruction of the Proto-Karaim Swadesh list. The main problem here is detection of the West Oghuz loans and especially of contact-induced archaization (fake archaisms) in Crimean Karaim. The objective of the second case study is to identify the genealogical affiliation of the Crimean Tatar dialects. Both the manual analysis of the innovations in the basic vocabulary and the computational lexicostatistics (Bayesian approach, Neighborjoining, Maximum Parsimony Analysis) confirm the traditional view that the Coastal dialect belongs to the Oghuz subgroup, the Orta dialect - to the West Kipchak subgroup, and the Steppe dialect - to the Nogai Kipchak subgroup. Such affiliations fully fit the documented ethnic history. The correct genealogical affiliation of the dialects in question became possible only after exclusion of all the loans, which has not been done in previous lexicostatistical studies of Crimean Tatar. Both cases show that careful elimination of areal influences is crucial for semantic (onomasiological) reconstruction and phylogenetic studies.

Keywords: phylogeny; semantic reconstruction; lexical borrowings; Karaim language; Crimean Tatar language; Turkic languages.



Аннотация

И.М. Егоров. Базисная лексика близкородственных языков в ситуации языкового контакта: тюркские языки Крымского полуострова

Настоящая статья объединяет два разыскания в области базисной лексики тюркских языков Крымского полуострова. Ее цель -- заострить внимание на проблемах, с которыми сталкиваются лингвисты при диахроническом -- и в особенности филогенетическом -- анализе интенсивно контактирующих друг с другом близкородственных языков. Первое исследование посвящено ономасиологической реконструкции пракараим- ского списка Сводеша. Основная рассматриваемая здесь проблема -- выявление запад- но-огузских заимствований и, в первую очередь, контактно обусловленных архаизмов (fake archaisms) в крымском диалекте караимского. Задача второго исследования -- определение генеалогической принадлежности крымскотатарских диалектов. Ручной анализ инноваций в базисной лексике и алгоритмы вычислительной филогенетики (байесовский метод, метод ближайших соседей, метод максимальной бережливости) подтверждают традиционное мнение о том, что береговой диалект принадлежит к огузской группе, средний -- к западно-кыпчакской, а восточный -- к ногайско-кып- чакской группе. Такой результат полностью подтверждается данными по этнической истории. Установить правильную генеалогическую аффилиацию рассматриваемых диалектов удалось только после выявления всех заимствований, чего не делалось в предыдущих лексикостатистических исследованиях по крымско-татарскому языку. Оба изученных кейса показывают, что элиминация ареальных влияний принципиально важна и для семантической (ономасиологической) реконструкции, и для филогенетических исследований.

Ключевые слова: филогенетика; семантическая реконструкция; заимствования; караимский язык; крымскотатарский язык; тюркские языки.

Introduction

The procedure of reconstruction in comparative-historical linguistics implies being able to distinguish between inherited and loaned items and patterns. This statement is true for phonologic, morphologic, and semantic reconstruction. Usually, when a word violates regular sound correspondences, it is treated as a borrowing unless it can be explained as the result of an analogical or another occasional change. Of course, extra sets of sound correspondences can appear between remotely related or unrelated languages as a result of phonological adaptation as well. However, as a rule, such borrowings can be revealed relatively simply, based on their distribution in the contacting subgroup. Various specific problems arise in the case of borrowings from a genetically related language, cf. for example the so-called “etymological nativisa- tion”, described in detail by Ante Aikio (2007) for Finnish loans in the Northern Saami.

Problems caused by contacts between closely related languages are relevant not only for traditional historical-linguistic studies, but also for linguistic phylogeny. The issue of homo- plasy and especially horizontal transfer has been redefined in the last decades, see Nakhleh, Ringe & Warnow 2005; Nelson-Sathi et al. 2011 and Kassian 2017. These works make linguists aware of the problem and propose methods to uncover and eliminate it. Early criticisms of Moris Swadesh's lexicostatistic and glottochronological methods were caused mostly by incorrect interpretation of loans. One of the most known critical works is Knut Bergsland and Hans Vogt's paper (1962), where it was argued that literary Norwegian (Riksmвl) demonstrates a drastically longer distance from Old Norse than Icelandic. The problem was that both contact- induced and autonomous replacements in the basic vocabulary were considered valid for measuring genealogical distance, whereas in reality the effect of the first group is highly dependent on the specific sociolinguistic situation. Revisiting this case, Sergei Starostin (2000: 230) has shown that 16 of 20 innovations in Riksmвl are loans: 11 from Danish, 3 from Swedish and 2 from German. Hence, if they are excluded, the percentage of innovations more or less equals that in the other Scandinavian languages. Nowadays, the detection of loans when reconstructing phylogeny has become an obligatory requirement at least in the Moscow school of comparative linguistics.

However, such drawbacks still arise in more recent phylogenetic studies applying lexico- statistics. For instance, confounding true cognates and borrowings, Russell Gray and Quentin Atkinson (2003) and then Remco Bouckaert et al. (2012) have inferred such a structure for the Slavic group in which Polish forms one clade with Belarusian, Ukrainian, and Russian. This contradicts the existing consensus which assumes a trifurcation of Proto-Slavic into the following subgroups: [Polish, Czech, Slovak, Sorbian], [Slovenian, Serbian/Croatian/Bosnian, Bulgarian, Macedonian], [Belarusian, Ukrainian, Russian]. Such affiliation of Polish is caused by undetected Polish loans in the Belarusian wordlist used in the forenamed works (see the linguistic supplement in Kushniarevich et al. 2015 for more detailed criticism). In section 4.6, I address identical problems in a recent work on Turkic phylogeny.

In the present paper, I intend to discuss two cases which illustrate the problems with the basic vocabulary of the languages undergoing intensive influence on the part of their close relatives. In my investigation of the Turkic languages of Crimea, I attempt to show the challenges they pose to a historical linguist, when the new method of onomasiological reconstruction is applied to identify the genealogical affiliation of a language. The Turkic languages spoken on the Crimean Peninsula provide suitable material for discussion of these issues for the following reasons: (a) they are related to each other approximately at the same depth as Riksmвl and its Scandinavian relatives from the canonical example cited above; (b) the tree structure and historical phonology of the Turkic family are known well enough for the purposes of our research; (c) the ethnic, sociolinguistic, and political history of Crimea is well documented.

The remainder of this paper is structured in the following way. Section 2 contains basic information on the Turkic languages of Crimea, their traditional genealogical affiliation, and the sociolinguistic situation in the region, along with a short annotated bibliography. Sections 3 and 4 deal with semantic (onomasiological) reconstruction of the Proto-Karaim wordlist and with revision of the genealogical affiliation of the Turkic varieties spoken in Crimea respectively. Each section contains its own introductory, methodological, analytical subsections and discussions of the results. Section 5 summarizes what can be learned from the considered cases.



1. Turkic languages of Crimea

In this section I provide the most important information on the sociolinguistic situation in the Crimean Peninsula, traditional genealogical affiliation of the languages, dictionaries, grammar and other sources used in the present study.

Crimean Tatar

The group of dialects traditionally referred to as the Crimean Tatar language actually represents a paraphyletic formation (Sevortyan 1966). It includes (1) Coastal dialect, which is genetically an Oghuz language most closely related to Turkish and Gagauz, (2) Orta (also called

Figure 1. Turkic varieties of Crimea. The map has been drawn on the basis of the Soviet ethnographic map of Crimea 1926; Filonenko 1931 and Radloff 1896: xiv-xvi. The border between the Coastal and Orta dialects is somewhat arbitrary

Central or Middle) and (3) Steppe dialects, both belonging to different Kipchak subgroups. The Coastal dialect is sometimes named Crimean Turkish; such term reflects its genealogical affiliation exactly. It became the dominant language in the Crimean Khanate, which was a vassal state of the Ottoman Empire. Modern literary Crimean Tatar is based on the Orta dialect.

Dictionaries: Useinov 2007 - dictionary of the Literary Crimean Tatar which is based on the Middle dialect.

Grammars: Sevortyan 1966; Izidinova 1996 - short grammar sketches; Kavitskaya 2010 - grammar based on the field notes from the early 2000s.

Other materials and studies: Polinsky 1992 - 100-wordlists for three Crimean Tatar dialects. I also use the wordlist recently collected after my own initiative, which can be found in Supplement 1.

Karaim

Karaim (also called Karaite) is a subgroup of Kipchak languages consisting of three dialects, sometimes treated as three separate languages. Only one of them was spoken by Jewish Karaite community in Crimea until recently. Two other dialects (Trakai and Halich) appeared as the result of the migration of Karaims from Crimean Khanate to the Grand Duchy of Lithuania. The migration to Trakai started in 1397, to Halich in 1407-1409 and continued up to the 15^th and 16^th centuries (Musaev 2010: 205-206). Karaim is traditionally classified together with Karachay-Balkar, Kumyk and with the Middle dialect of Crimean Tatar as a West Kipchak language (Johanson 1998: 82). Maria Polinsky treats Crimean Karaim as an “ethnolect” of Crimean Tatar belonging to the Oghuz subgroup.

Dictionaries: Baskakov, Szapszal & Zaj^czkowski 1974 remains the most reputable source on the lexicon of all dialects; Aqtay & Jankowski 2015 deals with Crimean Karaim, includes all Crimean materials from Baskakov, Szapszal & Zaj^czkowski 1974 and from other written sources.

Grammars: Musaev 1964 deals with the Trakai and Halich dialect; Musaev 2010 contains information on Crimean Karaim as well; Prik 1976 - grammar sketch of the Crimean Karaim.

Other materials and studies: Kocaoglu 2006 - texts in the Trakai dialect with brief grammar sketch and vocabulary; Polinsky 1992 - 100-wordlist for the Crimean dialect; 110-wordlist for Trakai dialect, speaker's self-recording made in 2019.

Krymchak

Krymchak is the language of the other Jewish community, which survived until the end of 20^th century in the town of Qarasuvbazar (Ukr. Bilohirsk). In the late 20^th and early 21^st century, some attempts at revitalization were undertaken. However, now this language is extinct. In a number of works, Krymchak is treated as a Kipchak language. Polinsky names it (as well as Crimean Karaim) an “ethnolect” of Crimean Tatar, i.e. an Oghuz language.

Dictionaries: Rebi 2004 - the dictionary created by language activists; Ianbay 2016.

Other materials and studies: Polinsky 1992 - 100-item wordlist and short grammar sketch; Polinsky 1991 - text sample; Jankowski 2017 - overview of grammar and major sources.

2. Reconstructing the Swadesh wordlist for Proto-Karaim

Introductory remarks

Traditional reconstruction of lexical semantics remains extremely arbitrary. A typical meaning of a reconstructed root or even a lexeme is `a kind of tree' or `to stack, to collect, to dump, to put in order, to build up'. Such definitions are the results of two wrong methods of semantic reconstruction: 1) reduction of all meanings attested in the daughter languages to a wide semantic component; 2) extrapolation of all attested meanings onto the proto-language. Sometimes this results in openly ridiculous situations: thus, according to Dybo 1996: 18, about 70% of the Proto-Indo-European verbal roots in Julius Pokorny's dictionary (1959) mean `to bloat, to swell' or `to bend'. For further criticism of the traditional semantic reconstructions see Bur- lak & Starostin 2005: 248. To solve this problem, the method of onomasiological reconstruction has been elaborated in the recent years. I will discuss it in the next section.

Why do we need reconstructed Swadesh lists? It seems reasonable to use reconstructed wordlists for commonly accepted low-level groups when investigating the tree structure of a deeper family. The principle of step-by-step reconstruction is a commonly accepted standard in comparative-historical studies. It is widely applied for phonological reconstruction. For instance, if one introduces a Germanic word into Indo-European comparison, methodologically it is more correct to use a reconstructed Proto-Germanic form instead of Gothic, Old High German, Old North and Old English, since each of them demonstrates innovations that are irrelevant to external comparison. Similarly, more correct is the use of a Proto-Germanic word- list when reconstructing the Indo-European tree. This exact approach was recently used in Kassian et al. forthcoming. The reconstructed Proto-Karaim wordlist can be used when revising the topology of Turkic family and reconstructing the Proto-Turkic wordlist for further comparison.

Methods

A relatively strict method of onomasiological reconstruction was recently developed by the representatives of the Moscow School of comparative linguistics (see Kassian, Starostin & Zhivlov 2015: 304-306; Starostin 2016). It involves tracing a way from the meaning to its optimal exponent in the protolanguage, i.e., determining which word was used for a given concept in a protolanguage. Selection of the optimal candidate is guided by five principles, which are very similar to the ones used for detecting archaisms and innovations when reconstructing phonology. The basic principle is topological (1); others (2-5) are used in competitive situations, i.e. when tree topology allows no unambiguous judgment on the candidates. Here I only provide a brief synopsis; for strict definitions, further explanations and examples see the abovementioned works:

1) tree topology: the root attested in different branches is preferable;

2) external etymology: the root is preferable if its external cognates preserve the same Swadesh meaning;

3) internal derivability: the primary root (as opposed to polymorphemic derivates) is a preferable candidate;

4) typology of semantic shifts: the typologically frequent direction of semantic shifts to be assumed when reconstructing scenario of semantic changes of the potential candidates;

5) areal effect exclusion: contact-induced innovations to be excluded.

It must be noted that these principles are different from the criteria for synchronic word- lists: terms for synchronic Swadesh wordlists are selected based on its frequency and stylistic neutrality.

Another advantage of onomasiological reconstruction is that it allows the elimination of uncertainties and mistakes in a synchronic list. Thus, if the quality of data on one language is not irreproachable, this can be compensated for with data on its relatives. In the process of reconstruction, an incorrectly selected term for one of the languages is likely to be seen as an innovation in this particular idiom. If a wordlist is overcrowded with inappropriate archaisms, the situation becomes more difficult. The solution to this problem is proposed in Section 3.3. Although the probability that one inaccurate list will influence the structure of the phylogenetic tree is lower and onomasiological reconstruction helps fix some defects in the data, the motto “garbage in - garbage out” remains fully true.

Next, I will concentrate on detecting contact innovations in Crimean Karaim, which make relatively shallow Proto-Karaim reconstruction difficult. Since the phonological inventory of Crimean Karaim and Coastal Crimean Tatar is very similar to each other, the phonological criterion is not particularly helpful in this case. Obviously, the reflexes of Proto-Turkic stems differ in these languages and sometimes these differences point to the Oghuz origin of the word. However, this criterion cannot be applied to every word. I will mostly use the distributional criterion, which can be described as follows. For instance, four languages (L1, L2, L3, L4) related to each other among which L1 is an outgroup and L2, L3, L4 form a separate clade are taken into consideration. The lexeme A is the basic term for the meaning `M' in L1. In L2, lexemes A and B are synonyms; in L3, L4, `M' is denoted only by the lexeme B (see Figure 2). If A is a primary root whereas B is a transparent derivative or semantic innovation, this constitutes strong evidence for reconstructing *A for the meaning `M' in the Proto-L2-4. However, if L1 influences L2, A should be regarded as a borrowing. The coexistence of A and B in L2 is an additional argument for such a solution.

Figure 2. Tree structure and distribution of lexemes with meaning `M' which hint at horizontal transfer of the stem A

Extrapolating this scheme to the case in question, I will regard a word as an Oghuz borrowing in Karaim if it is widely spread in the Oghuz or at least in the West Oghuz languages and is uncommon or completely absent in Kipchak.

In addition to being subject to horizontal transfer (MAT-borrowings in Jeanette Sakel's (2007: 15) terminology), some stems can also undergo contact-induced semantic shifts (PAT- borrowings in Sakel's terminology). The last phenomenon is also known as loan meaning extensions (see Haspelmath 2009). In the case of closely related languages, this is driven by obvious, naпve logic: “if words sound similarly they must have similar meaning”.

Contact innovations in Crimean Karaim: clear cases

More evident borrowings will be considered at first. All of them are loaned from the Coastal dialect of Crimean Tatar. Some of these loans were in turn borrowed into West Oghuz languages from Persian and Arabic. Theoretically, one can assume that Oghuz-like lexemes in Crimean Karaim are inherited and Kipchak-like ones are borrowed. However, such assumption faces more difficulties, since it is hardly possible to find the source of potential Kipchak borrowings which occurred in the Trakai and Halich dialects and sporadically in the Crimean dialect as well.

In citing examples, I first give the number of the concept in the 110-item Swadesh list following Kassian et al. 2010, then list the language material with references. The abbreviation for the source used in Baskakov, Szapszai & Zajqczkowski 1974 and Aqtay & Jankowski 2015 is given in brackets for Crimean Karaim forms. For the full reconstructed Proto-Karaim Swadesh wordlist see Supplement 1. For the transcription and transliteration of the examples, I use the Unified Transcription System applied in the Global Lexicostatistical Database (https://starling.rinet.ru/new100/UTS.htm).

5. big - CrKar. balaban (Sz) `big, huge' (Baskakov, Szapszai & Zajqczkowski 1974: 100). It is difficult to define whether balaban is appropriate even for the synchronic Crimean Karaim Swadesh list. There are two other candidates that will be considered below. CrKar. balaban is a clear Oghuz borrowing, cf. Tur. balaban `huge', Gag. balaban `high', CoCrTat. balaban `big'. The root is extremely rare beyond the Oghuz languages. Only two comparanda are mentioned in Dybo 2013: 128-131: Tat. dial. balban `fat, overweight, stout', Kirg. balpay- `to seem big, bulky, clumsy'. Details of its etymology remain obscure (see the cited work for the review of existing hypotheses), however, the innovative nature of the meaning `big, huge' is quite obvious.

CrKar., TrKar., HKar. biyik (Sz, Par. 101 v. 1) `big, high, great' (Baskakov, Szapszai & Za- jqczkowski 1974: 115) is another stem which could be treated as a contact-induced innovation. This stem with the meaning `high' is widespread across Turkic languages (see Dybo 2013: 123) whereas the more general meaning `big' is limited to the Oghuz subgroup (Tur. bьyьk, Gag. bьk, Az. bцyьk), Karakhanid Uyghur, Old Uyghur, and Sary Yugur. Such distribution theoretically can be an indication of the antiquity of the meaning `big' (cf. Clauson 1972: 302 for an interpretation), but *ulug is the better candidate for both Proto-Karaim and Proto-Turkic `big'. All Karaim dialects demonstrate its reflexes: CrKar. ulu (Sz) ~ uli `great, big', TrKar. ullu `big, great, important' ~ urjlu `great, big, elder' ~ ullux `big, great' (with additional suffix), HKar. ullu `big, great, important' (Baskakov, Szapszaf & Zajqczkowski 1974: 577, 579). At least modern speakers of the Trakai dialect use ullu as the basic word for `big', according to our data; it is also confirmed with materials published in Kocaoglu 2006. The semantic shift `big' > `great' is typical for the world's languages, the synchronic polysemy is also common in the Turkic family (Dybo 2013: 120-121) and cross-linguistically (Rzymski et al. 2019). The direction of the shift `big (of a physical object)' > `great (high status)' is more probable than vice versa due to the common tendency of the development from concrete meanings to more abstract ones (Campbell 2013: 237). In sum, the old term for `big' in the Proto-Karaim subgroup is *ullu; the semantic shift `high' > `big' of the stem *biyik should have been triggered by contact with the Oghuz dialect of Crimea before the start of the migration to the Grand Duchy of Lithuania. Note also the form bьyьk in the Crimean dialect, which looks as if it was recently borrowed from Turkish. Aqtay and Jankowski (2015: 88, 100) list the latter form with the gloss `great, big' whereas biyik is glossed as `high'.

It is difficult to make a choice between three candidates with the meaning `big' for synchronic Crimean Karaim based on existing sources, which partly contradict each other. Provisionally, I assume that ulu ~ uli has the more abstract meaning `great' whereas biyik ~ bьyьk and balaban compete with each other in the basic meaning `big'. Both are loans, the meaning `big' by the lexeme biyik is borrowed from Oghuz; bьyьk and balaban are MAT-borrowings. The choice is much simpler for Proto-Karaim. Thus, this case illustrates an important advantage of onomasiological reconstruction: uncertainty in the data on one of the languages does not influence the final list.

67. red - CrKar. qirmizi ~ qirimzi `red'. This is an Arabic loan common in Oghuz languages, cf. Tur., Gag. qirmizi `red', Az., Turkm. Girmizi `red'. In the Trakai and Halich dialects, kirmizi denotes a specific shade `purple, magenta' (Baskakov, Szapszal & Zajqczkowski 1974: 381, 387). The archaic stem kizil, reflecting PT *Kir^yil `red', has been found with the meaning `red, orange' for these dialects; it should be the basic term for `red' (Baskakov, Szapszal & Za- jqczkowski 1974: 383). The reflexes of *КігУіі are not attested for Crimean Karaim in Baskakov, Szapszal & Zajqczkowski 1974, but Aqtay and Jankowski (2015: 309) cite it with the gloss `red, ruddy'. The stem *qizil can be reconstructed for Proto-Karaim `red' with complete certainty. It is one of the most stable Turkic stems.

68. root - CrKar. kцk `root' (Baskakov, Szapszal & Zajqczkowski 1974: 337). This stem is widely attested with the meaning `root', however, its basic meaning in the majority of the Turkic languages is more abstract (`basis'), it develops various metaphorical meanings as well. As the basic term for `root' this stem is attested in the Oghuz languages, cf. Tur., Gag. kцk, from which it has been borrowed into Crimean Karaim. The stem *damor > *tamur is a better candidate for Proto-Turkic and Proto-Karaim `root'. It is preserved all over the Turkicspeaking area including the languages in question and their numerous Kipchak relatives, cf. Karaim reflexes: CrKar., tamur ~ tamar `root, vein', HKar. tamar ~ tamur `root, vein', TrKar tamur `root, vein' (Baskakov, Szapszal & Zajqczkowski 1974: 509-510).

69. round - CrKar. mьdever ~ mudever `round', yuvarlaq `round, globular', tomalaq `round, full, plump', yumalaq `globular, round' (Aqtay & Jankowski 2015: 247, 407, 472, 474), tцgerek `round' (Baskakov, Szapszal & Zajqczkowski 1974: 541). The first term has an Arabic origin and is borrowed via Turkish, cf. Tur. mьdevver `circular, round'. The second one is an Oghuz borrowing as well, cf. Tur., Gag. yuvarlaq `round'. The stems tomalaq and yumalaq have been attested in other Turkic languages:

Uzb. yumaloq `round (sphere & circle)', Karak. zumalaq `round (circle)', for other derivatives with the Proto-Turkic bound root *yum- see Dybo 2013: 441-442;

Gag. tombarlaq `round', Uzb. dumaloq `round (sphere & cylinder)', Uyg. domlaq `round (sphere)', Bash. tumalaq `round (sphere & circle)', Nog. timalaq `circle, sphere (n.)', Kaz. domalaq `sphere (n.)', Karak. dumalaq `round (sphere)' - note that this set of phonetically similar forms demonstrates suspiciously irregular sound correspondences!

However, none of the Karaim stems listed above have been sufficiently confirmed by the Trakai and Halich data; the meaning `round' is insufficiently documented in the existing sources. Only the stem tцgerek has a Trakai cognate. According to the recently collected word- list for the Trakai dialect, either the collocation galgan kibik literally means `circle-like' or tдgere ~ tegдrдk `round, circle' is used as an adjective `round'. CrKar. tцgerek and TrKar. tдgere ~ tegдrдk are treated as a Mongolian reborrowing and an inherited stem respectively in Dybo 2013: 238-239. However, I believe that it is reasonable to consider them true cognates and to reconstruct * tцgerek for Proto-Karaim with the meaning `round (circle)'.

Until there is a corpus for Karaim, the choice for the synchronic basic term is difficult both for the Crimean and Trakai dialects. The stem tдgere ~ tegдrдk is a single candidate for Trakai `round', Crimean tцgerek can apply at least for `round 2D'. Further details in synchronic dialects remain obscure.

79. smoke - CrKar., TrKar. tьt-sь (Sz) `smoke, incense' (Baskakov, Szapszal & Za- j^czkowski 1974: 555). This is an old contact innovation shared by the Crimean and Trakai dialects. The substantives from the verb PT *tьt- formed with the not especially productive suffix *-sьg have been found only in Oghuz languages, cf. tьtsь Tur. `incense', Az. tьstь `smoke' with metathesis inside the consonant cluster, Turkm. tьsse `smoke', Sal. tissi `smoke', for this suffix see Rдsдnen 1957: 141. All other Turkic languages, including even Chuvash, demonstrate the suffix *-ьn (Dybo 2013: 479). So we treat CrKar. and TrKar. tьt-sь `smoke, incense' as a borrowing which occurred before the migration from Crimea to the Grand Duchy of Lithuania. The inherited forms with *-ьn have been found in all Karaim dialects as well: CrKar. tьtьn `smoke, tobacco', TrKar. t<sup>y</sup>ut<sup>y</sup>un<sup>y</sup> `smoke, tobacco', HKar. titin `smoke' (Baskakov, Szapszal & Za- j^czkowski 1974: 532, 555, 571). Hence, *tьtьn must be reconstructed for Proto-Karaim `smoke'.

103. near - CrKar. yaqin is attested in the Crimean dialect (Baskakov, Szapszal & Za- j^czkowski 1974: 220) beside CrKar. yuwuq, TrCar. yuwux, HCar. yuwuk `near' (Baskakov, Szapszal & Zaj^czkowski 1974: 253-254). I consider the first stem an Oghuz borrowing: Tur. and Gag. yaqin `near', Az. yaxin `near', Turkm. yaqin `near'. Both stems are widely spread across Nuclear Turkic languages. However, the narrow distribution in the Karaim dialects allows us to treat *yaqin as a borrowing. Its competitor, *yuwuq, which can be found in all Karaim dialects, is definitely the better candidate for Proto-Karaim `near'. The distribution of its external cognates points to the stem discussed above as to the main exponent of the meaning `near' not only in Proto-Kipchak and even in Proto-Turkic, see Dybo 2013: 539-540.

Contact archaization in Crimean Karaim: fake archaisms

In this section, I consider the most curious cases. There are some stems which can seem archaic at first sight, but in reality turn out to be loanwords. For such cases, I suggest the term `fake archaisms'. Revealing this kind of borrowings is crucial for onomasiological reconstruction. Fake archaism must be suspected when principles of tree topology, external etymology, and internal derivability come in conflict with the principle of areal effect exclusion. The semantic plausibility principle, i.e. the typology of semantic shifts, theoretically, can also contradict the principle of areal effect exclusion but such cases have not been attested in our material.

Thus, fake archaisms can be successfully detected when areal distribution and the direction of influence are taken into account. In Crimean Karaim, four examples of fake archaisms have been found.

22. to eat - there are two candidates for filling this slot:

1) CrKar. ye- (Sz) `to eat' (Baskakov, Szapszat & Zajqczkowski 1974: 268);

2) CrKar., TrKar. as-a- (Sz) `to eat', HKar. as-a- `to eat' (Baskakov, Szapszat & Zajqcz- kowski 1974: 79, 91).

The root *ye- should have an advantage due to the principle of external etymology. It is found not only in numerous non-Kipchak languages but even in Chuvash, whereas *as-a- is limited to the Nuclear Turkic languages. In many of them, it is often a marked polite term = Rus. kusat'. The principle of internal derivability also speaks for the primary root *ye-, since the verbal stem *as-a- can be analyzed as a synchronic derivative from *as- `food'. However, *ye- is limited only to the Karaim dialect that was under intense influence on the part of Oghuz. The Oghuz languages preserve *ye- as the basic exponent of `to eat' (Tur., Az. ye-, Gag. i-, Turkm. iy-, Sal. yi-). Thus, one can simply consider CrKar. ye- a borrowing. A probable situation is that *as-a- already becomes the basic term for `to eat' in Proto-Karaim, but archaic *ye- as a marginal term still remains in Proto-Karaim. Under foreign influence *ye- could become the basic term again, i.e. we deal with a semantic backformation.

83. sun - three words glossed in this way have been found in Crimean Karaim, and two of them can apply for the status of the basic term in Proto-Karaim.

1) CrKar. kьn `sun, day' (Sz, R) ~ gun `day' (Par 77 v. 11), cf. TrKar. k^yun^y `day', HKar. kin `day' (Baskakov, Szapszat & Zajqczkowski 1974: 167, 320, 353, 396);

2) CrKar. kьnes ~ gьnes `sun' (Baskakov, Szapszat & Zajqczkowski 1974: 354; Aqtay & Jankowski 2015: 169, 225);

3) CrKar. quyas, TrKar. kuyas, HKar. kuyas `sun' (Baskakov, Szapszat & Zajqczkowski 1974: 344, 372).

The first item attested mostly with the meaning `day' demonstrates also the meaning `sun' in Crimean Karaim. It is a reflex of the stable Proto-Turkic stem *gьn `day, sun' which retained this meaning across the whole area of the Turkic languages. The second stem was derived from the first one with a not quite clear suffix. It occurs sporadically in various languages, cf. Tur., Kum. gьnes `sun', OT kьnes `sun'. The Crimean Karaim form kьnes ~ gьnes is a transparent Western Oghuz loan due to the initial voiced consonant. The last stem, quyas, is a result of the semantic shift `heat' > `sun', which should independently occur in a couple of Turkic languages. Thus, when one chooses between *kьn and *quyas, the external etymology principle strongly points to the first stem as the better candidate for filling the slot `sun' in Proto- Karaim. However, in light of Tur., Gag. gьn `sun, day' (attested simultaneously with gьnes in Turkish), it is reasonable to regard the meaning `sun' of CrKar. kьn as a result of backformation. Hence, the slot `sun' must be filled by the stem *quyas in Proto-Karaim. The retention of the stem *kьn in Halich and Trakai collocations kyuny batis `sunset' (lit. `sun diving'), kyuny tuvus `sunrise' (lit. `sun appearing') can prove its antiquity in this meaning and, hence, the existence of *kьn `sun' in Pre-Proto-Karaim, but it remains questionable whether this evidence is sufficient to reconstruct Proto-Karaim *kьn as the basic term for `sun'.

84. to swim - two candidates for this slot have been found:

1) CrKar. yьz- (Sz) ~ ьz- (Sz) `to swim' (Baskakov, Szapszat & Zajqczkowski 1974: 261, 588);

2) CrKar. com- (Sz) `to swim, to dip', TrKar. com- `to swim, to dip', HKar. com- `to swim, to flow' (Baskakov, Szapszat & Zajqczkowski 1974: 614, 632, 639).

The first one reflects the relatively stable Proto-Turkic stem *yьU- `to swim', cf. Oghuz comparanda: Tur., Turkm. yьz- `to swim', Gag., Az. ьz- `to swim', it is also common beyond the Oghuz subgroup (Dybo 2013: 490). This stem can be safely reconstructed for Proto-Turkic `to swim'. The second candidate is a transparent innovation. It reflects the semantic shift `to dive, to dip' > `to swim'. The original meaning is confirmed by a number of languages:

Tuv. sim-in- (refl.) `to dip, to dive', OUyg., KarakhUyg., Chag., Uyg. com- `to dip, to dive', Uzb. com- `to dip, to dive', Tat. cum- `to dip, to dive', Chuv. cem- `to dip, to dive' (Tat. borrowing?), Bash. sumi- `to dip, to dive' (Dybo 2013: 491).

Note that the polysemy `to swim, to dive' is attested in the Karaim dialects as well. Based on the external etymology principle, one could reconstruct *yьz- for Proto-Karaim `to swim'. However, this stem must be regarded as a borrowing since it is limited to the one dialect in intimate contact with West Oghuz, while the archaic stem, on the contrary, is retained in Karaim.

86. that / 87. this - the system of demonstrative pronouns in the Crimean Karaim has been influenced on the part of Oghuz languages.

Table 1. The subsystems of the demonstrative pronouns in the Karaim languages

	Crimean	Trakai	Halich
proximal	bu `this'	bu `this'	uspu `this here'	bu `this'	uspu `this here'
medial	su `this, that'	--	--	--	--
distal	ol `that'	ol that'	osol `that there'	ol that'	osol `that there'

The three-way deictic system, like in Crimean Karaim, can be potentially treated as archaic. Proto-Nuclear-Turkic *su functions as a medial deictic pronoun in several languages: Gag. su `this, that (medial deixis)', Turkm. su `this, that (medial deixis)', Uzb. su `this, that (medial deixis)', Kum. su `this, that (medial deixis)', Kirg. su `this, that (medial deixis)'.

Theoretically this could confirm the antiquity of CrKar. su. The systems with bare *su as a medial deictic pronoun are common in the Oghuz languages (Tenishev & Dybo 2002: 145156), but not typical for other Turkic subgroups. Outside Oghuz, the Proto-Turkic pronominal root *su is more frequently attested with various extensions:

Chuv. fav3, /акз `this', les `that', Yak. sol `that', Turkm. sol `that', Bash. oso `this', sul `that', Tat. sul `that', Nog. sosi `this', sol `that', Kaz. osi `this', sol `that', Karak. usi `this', sol `that', Kir. usu `that'.

The fact that Crimean Karaim, Kumyk, and Kirgiz feature simply *su sets them apart from other Kipchak languages. Therefore, it may be suspected that Crimean Karaim demonstrates another Oghuz loan. Thus, only *bu `this' and *ol `that' can be reconstructed for Proto-Karaim with certainty. In fact, *su must not be a deictic pronoun but rather a deictic particle, see Dybo 2013: 497-498.

Phonological variation in Crimean Karaim

Another result of strong Oghuz influence on Crimean Karaim is the presence of phonological doublets which reflect both Kipchak and Oghuz development of the same Proto-Turkic root. The Oghuz-like counterparts are borrowings. Due to the fact that Oghuz looks more archaic than Kipchak in some parameters, these cases, considered in Sections 3.4.1 and 3.4.2, can also be regarded as fake archaisms.

Reflexes of PT *g

To the basic distinctions between Oghuz and Kipchak languages belong the reflexes of *g after a low central vowel. The Oghuz languages demonstrate an uvular consonant whereas the west majority of the Kipchak languages change the velar to a labial. A school-book example is the reflex of Proto-Tukic *da:g `mountain':

Oghuz: Tur. da: (dial. daw), Az. daw, Turkm. da:G, Sal. da:K;

Kipchak: Kum. taw, K.-B. taw, Tat. taw, Bash. tau, Nog. taw, Kaz. taw.

Crimean Karaim demonstrates both taw `forest' (Sz) and taw `mountain' (Sz); the third variant is daw `mountain' (ZR 45, 3). These forms contrast with TrKar., HKar. taw `mountain' (Baskakov, Szapszat & Zajqczkowski 1974: 168, 503, 505). The Crimean Karaim form taw must be treated as phonologically adopted. Voiced d was substituted with voiceless t, since only voiceless dentals are possible in word onset in the inherited vocabulary. The final velar does not undergo the adaptation since there is no general restriction on w after vowels at least in the non-final position, cf. aljawim `I will take', qartnip tayawi `old man's stick'. A simultaneous occurrence of adopted (to various degrees) and non-adopted items is typical for the situation of intensive influence, cf. Russian loans in Kazym Khanty and Finish loans in Northern Saami:

Khant. askola ~ skola `school' < Rus. skola `school';

Khant. wentdr ~ andrey ~ andr^yey `a male personal name' < Rus. Andrey `a male personal name';

SaaN. hirbmat ~ harbmat `horrible' < Fin. hirmu `horror' (Aikio 2007: 28-29);

SaaN. hapmu `craving (for a particular food)' ~ hipmu `lust, desire' < Fin. himo `carving, desire' (Aikio 2007: 28-29).

Another example for *-ag in the final position found in the Karaim Swadesh list:

CrKar. yaw (Sz) ~ yaw (Sz) `fat', cf. TrKar., HKar. yaw `fat' (Baskakov, Szapszal & Za- jqczkowski 1974: 214-215).

Reflexes of the Proto-Turkic vocalic-consonantal cluster *-agi- are a special case. In Kipchak, not only does *g become a labial consonant, but *i also becomes a rounded vowel. Oghuz demonstrates here an uvular consonant and an unrounded vowel.

CrKar. awur (Sz) ~ awir (Sz, R) `heavy', cf. TrKar., HKar. awur `heavy' (Baskakov, Szapszal & Zajqczkowski 1974: 42, 44);

CrKar. awuz (Sz) ~ awiz (Sz) ~ awiz (Par 84 v. 9) `mouth', cf. TrKar., HKar. awuz `mouth' (Baskakov, Szapszal & Zajqczkowski 1974: 42, 44);

CrKar. bawur `liver' (Sz) ~ bawir (Sz) `chest, liver', cf. TrKar., HKar. bawur `liver', TrKar. bawir `liver' (Baskakov, Szapszal & Zajqczkowski 1974: 94, 96).

In Tenishev & Dybo 2006: 72-73, the double reflexes of *ag# and *agi have been postulated for Karaim, i.e. aw ~ ay and awu ~ ayi. It seems more reasonable to regard the reflexes with у as a result of Oghuz influence. If they are eliminated, Karaim will not differ from other Kipchak languages in its reflexes of *ag# and *agi. In the opposite case, the Karaim data would require reconstructing velar (or rather uvular) consonants for Proto-Kipchak in these clusters.

3.5.1. Initial voiced dental and velar consonants

Turkish and Gagauz reflect the Proto-Turkic distinction of initial voiced and voiceless dental and velar stops. For velars the opposition can be reconstructed only in roots with front vowels. The reconstruction of the initial Proto-Turkic voiced stops and some modifications which occurred in the Oghuz languages are described in all details in Tenishev & Dybo 2002: 68-83 (see also Dybo 2007 for further details and discussion). The majority of the Kipchak languages neutralize these oppositions in favor of the voiceless series. Crimean Tatar demonstrates contact- induced variation.

CrKar. keca `night' (Par 83 v. 3) ~ kece `night' (Sz, Man 3a, 8a) ~ geje `evening' (Kz III-IV, 81) ~ gece `night' (ZR 52, 26, Q 9) ~ geca `night' (ZR 52, 20), cf. TrKar. k<sup>y</sup>ec<sup>y</sup>a `night', HKar. kece `night' (Baskakov, Szapszal & Zajqczkowski 1974: 159, 167, 311-312, 394-395; Aqtay & Jankowski 2015: 164, 202) < PT *ge:ce;

CrKar. kel- `to come' (Sz, Cam, Dan 1:1, Man 2a) ~ gel- `to come' (Man 3a, Q34), cf. TrKar., HKar. kel- `to come' (Baskakov, Szapszal & Zajqczkowski 1974: 301-302, 390; Aqtay & Jankowski 2015: 164, 204) < PT *gel-;

CrKar. kцz `eye' (Sz, Cam, Psa 10:1, Man 5a) ~ gцz- `eye' (Par 82 v. 1, ZR 79, 15, ZR 95, 30, Man 5a, Q 4), cf. TrKar. k^yoz^y-, HKar. kez- `eye' (Baskakov, Szapszal & Zajqczkowski 1974: 161, 300, 312, 336; Aqtay & Jankowski 2015: 168, 221) < PT *gцr^y;

CrKar. kцr- `to see' (Par 83 v. 5, Man 1a, Q 38) ~ gцr `to see' (Q 36, 49), cf. TrKar. kyor-, HKar. ker- `to see' (Baskakov, Szapszal & Zajqczkowski 1974: 306, 314, 339; Aqtay & Jankowski 2015: 167-168, 218) < PT *gцr-;

CrKar. kьn `sun, day' (Sz, R, Man 3a) ~ gun `day' (Par 77 v. 11, Man 2b, Q 73), cf. TrKar. kyuny `day', HKar. kin `day' (Baskakov, Szapszal & Zajqczkowski 1974: 167, 320, 353, 396; Aqtay & Jankowski 2015: 169, 224) < PT *gьn;

CrKar. tas `stone' (Sz, Fil 7, 120, Q 81) ~ das `stone' (Par 83 v. 12, Q 21), cf. TrKar. tas `stone', HKar. tas `stone' (Baskakov, Szapszal & Zajqczkowski 1974: 170, 516, 518; Aqtay & Jankowski 2015: 132, 386) < PT *dia:A;

CrKar. taw ~ taw ~ daw < PT *da:g (details see above);

CrKar. tamar `vein, root' (Sz) ~ tamur `vein, root' (Sz, Q 431) ~ damar `vein' (ZR 78, 18), cf. TrKar., HKar. tamur `vein, root', HKar. tamar `vein, root' (Baskakov, Szapszal & Zajqczkowski 1974: 169, 509-510; Aqtay & Jankowski 2015: 131, 381) < PT *damor;

CrKar. terek `tree' (Man 10a) ~ teraq `tree' (Par 83 v. 14) ~ derek `tree' (Fil 8, 150, Q 58) ~ direk `tree' (Sz, ZR 44, 30) ~ diraq `post, column' (ZR 16, 21), cf. TrKar. tye^ak `fruit tree', HKar. terek `id' (Baskakov, Szapszal & Zajqczkowski 1974: 178, 185, 522, 565, 567; Aqtay & Jankowski 2015: 136, 396) < PT *derek;

CrKar. tьz `knee' (Sz) ~ diz `knee' (Q 628) ~ dьz `knee' (KM 61b), cf. TrKar. tiz ~ tz `knee', HKar. tiz ~ kiz `knee' (Baskakov, Szapszal & Zajqczkowski 1974: 317, 525-526; Aqtay & Jankowski 2015: 139, 144) < PT *di:ry;

CrKar. toli `full' (Sz, Par 102 v. 13) ~ tolu `full' (Sz, R) ~ doli `full' (Q 187), cf. TrKar., HKar. tolu `full' (Baskakov, Szapszal & Zajqczkowski 1974: 537; Aqtay & Jankowski 2015: 140, 407) < PT *do:l-;

CrKar. til `tongue' (Q 18, 49, Meq 60, 70) ~ dil `tongue' (Q 223), cf. TrKar. tiP, HKar. til ~ kil `tongue' (Baskakov, Szapszal & Zajqczkowski 1974: 319, 528; Aqtay & Jankowski 2015: 403, 138) < PT *dil ~ *dil;

CrKar. tis `tooth' (Sz) ~ cis `tooth' (Sz, Q 125) ~ dis `tooth' (ZR 70, 12, Q 302); cf. TrKar. tis `tooth', HKar. tis ~ kis `tooth' (Baskakov, Szapszal & Zajqczkowski 1974: 178, 323, 531-532, 629; Aqtay & Jankowski 2015: 124, 132, 404) < PT *di:A;

CrKar. tur- `to stand' (Sz, Man 5a) ~ dur- `to stand' (Par 77 v. 12, Q 54), cf. TrKar., HKar. tur- `to stand' (Baskakov, Szapszal & Zajqczkowski 1974: 181, 547; Aqtay & Jankowski 2015: 142, 413) < PT *dur-.

It should be noted that not all Proto-Turkic stems with initial *d found in Swadesh list demonstrate voiced consonants in Crimean Karaim: PT *dirpa-k > CrKar. tirnaq `fingernail, claw', PT *di:A-le- > CrKar. tisle- ~ cisle- `to bite', *deri > CrKar. teri `skin', *dury > CrKar. tuz `salt'. Such inconsistency indicates that in this case they are not regular reflexes but borrowings. The Proto-Turkic stems with initial *k and *t are found always with voiceless consonants: PT *kьl > CrKar. kьl `ashes', PT *kцn- > CrKar. kьy- `to burn (intr.)', PT *kцp > CrKar. kцp `many', PT *kiAi > CrKar. kisi `man (person)', PT *kicьk > CrKar. kici `small', PT *kem > CrKar. kim `who', PT *tь:k > CrKar. tьk `feather', PT *tьn > CrKar. tьn `night', PT *tьt-ьn > CrKar. tьtьn `smoke'.

Other Oghuz loans

The initial consonant of PT *s(i)ac `hair' yields c or s in majority of Turkic languages. However, reconstruction of the initial *s is proven by Yakut as (where *s- > 0- regularly) and Oghuz reflexes with retained s-. Crimean Karaim demonstrates doublets with Oghuz- and Kipchak-like reflexes. The first one should be a loan, since the Halich and Trakai dialects point to Proto- Karaim *c. This case belongs to fake phonological archaisms.

CrKar. sac (Par 107 v. 13) `a hair (Rus. volos - hair[SG])' ~ sec (Sz) `hair (Rus. volosy - hair- pl); tuft, crest' ~ cac (R) `a hair, hair', TrKar. cac `a hair, hair', HKar. cac `hair, fur' (Baskakov, Szapszal & Zajqczkowski 1974: 470, 500, 613, 625).

Two more Oghuz loanwords in Crimean Karaim are ver- `to give' and var- `to go', which reflect an Oghuz innovation. Although these words are not fake archaisms, I include them here since they additionally confirm the direction of borrowings in the pair Crimean Karaim < Coastal Crimean Tatar / Turkish. These stems reflect the shift of initial *b to v in monosyllabic stems with r in the coda.

CrKar. ber- (Sz, R) ~ ver- (Par 77 v. 19, ZR 32, 27) `to give', cf. TrKar. b<sup>y</sup>er- `to give', HKar. ber- `to give' (Baskakov, Szapszal & Zajqczkowski 1974: 112, 151, 158);

CrKar. var- (Q 4) ~ bar- (Sz) `to go', TrKar., HKar. bar- `to go' (Baskakov, Szapszal & Za- jqczkowski 1974: 102; Aqtay & Jankowski 2015: 436).

Preliminary conclusions

The onomasiological reconstruction of the Proto-Karaim Swadesh list is complicated mainly by the set of fake archaisms. Fake archaisms are a particular type of homoplastic development, namely MAT-borrowings and semantic back-formations from a sister subgroup which preserved more archaic (in the perspective of a whole family) items. In the Crimean Karaim case, Oghuz nature of the archaic-looking items is proven by the large amount of other Oghuz borrowings and by the history of the sociolinguistic situation.

Examination of sources used in Baskakov, Szapszal & Zajqczkowski 1974 and in Aqtay & Jankowski 2015 shows that some of them are more “Oghuzized” than others. From our data it is clear that Par, ZR, and Q contain many more Oghuz forms than Sz. Apparently, they demonstrate language shift to Coastal Crimea Tatar. Aqtay and Jankowski's more detailed study of the lexicon (2015: 9) confirms this statement. Data from these sources are inappropriate for phylogenetic studies.

Consistent detection of all borrowed elements allows mostly trivial reconstruction of the Swadesh list for the not particularly deep Proto-Karaim taxon.

3. Classifying languages of Crimea

In this section, I address the discussion of the genealogical affiliation of Crimean Karaim, each of three Crimean Tatar dialects and Krymchak. My goal within the scope of this section is not to build a complete phylogenetic tree but only to define the closest relatives of the idioms in question. Needless to say, disclosure of borrowings plays a crucial role in this procedure. Before comparing wordlists all loans, including inter-Turkic ones, must be excluded. Although this statement may seem trivial, in Section 4.6 it will be shown that even recent phylogenetic research still continues to be affected by undetected borrowings.

Previous research

Beginning with Radloff, the language of Crimean Karaims is fully identified with Crimean Tatar or seen as one of its dialects. This opinion is shared by Zajqczkowski, Doerfer, and Polinsky. At the same time, Crimean Karaim (whatever the term means) is included in the Karaim-Russian-Polish dictionary (Baskakov, Szapszal & Zajqczkowski 1974). In his earlier grammar (Musaev 1964: 36-37), Musaev maintains that the Crimean Karaits' variety is not distinguishable from Crimean Tatar and must not be included in the notion Karaim language. Information on the Crimean dialect was later included in his sketch of Karaim dialectology (Musaev 2010) by the editors. The discussion is summarized in the work by Jankowski (2003: 109-112), who attempts to show that Crimean Karaim is different from Crimean Tatar, involving phonological, syntactical, onomastic and lexical arguments. Basic vocabulary remains beyond his interest. The Swadesh list was examined in Polinsky 1992. She comes to the conclusion that Crimean Karaim together with Krymchak language is very close to the Orta and Coastal Crimean Tatar dialects and, hence, belongs to the Oghuz subgroup. Polinsky does not distinguish borrowed and inherited vocabulary when calculating lexicostatistical matches, therefore her conclusions can be called into question. To be fair, it must be noted that, to the best of my knowledge, the requirement to exclude contact innovations was yet to be explicitly formulated in 1992.

Currently, no detailed descriptions of the Crimean Tatar dialects exist and they are unlikely to appear in the future. Commonly accepted is Ervand Sevortyan's (1966) dialectal classification, which distinguishes three dialects of Crimean Tatar (Steppe, Coastal and Central) highlighting their heterogeneous origin. According to Sevortyan's classifications, the Steppe dialect belongs to the Nogai Kipchak subgroup; Orta is Cuman Kipchak, i.e. West Kip- chak; Coastal belongs to the Oghuz group. The original dialectal differentiation was violated as the result of Soviet deportation of Crimean Tatars to Uzbekistan in 1944. After the return to Crimea in the early 1990s, most families were not able to settle in their native villages. This provides further dialectal mixture. Already during Darya Kavitskaya's fieldwork in 2002-2003 and 2009, only older speakers had “clear dialect affiliation” (Kavitskaya 2010: 3). Dialectal mixture is quite visible both in my and Polinsky's data. See Normanskaya 2019 on dialectal mixture in literary Crimean Tatar.