1.2.1.1 Ancient Macedonian

Unlike modern Macedonian, which is a Slavic language, Ancient Macedonian – the language of the core troops of Alexander the Great (Katičić 1976: 106) – was an Indo-European language of uncertain connections within Indo-European. Katičić 1976: 108–112 identifies three layers in Macedonian vocabulary. One layer is clearly Greek and likely borrowed. Another layer has no Indo-European connection, and a third layer is clearly related to Greek but the words appear to be cognates rather than borrowings. The question, thus, is this: (1) Was Ancient Macedonian a Hellenic dialect distinct from the Hellenic dialect that gave rise to Ancient Greek (with all of its dialect diversity, on which see, e.g., Buck 1955), one whose non-Greek-looking vocabulary consists of loanwords, Greek words that did not survive in our known Greek sources (cf. Katičić 1976: 111), and reflections of inadequacies in the Greek transcription of Ancient Macedonian pronunciation, or (2) was it a sibling to Hellenic, i.e., part of a single pre-Hellenic diasystem that subsequently branched into Hellenic (whence the attested Greek dialects) and Ancient Macedonian, or (3) was it a non-Hellenic Indo-European language, possibly allied to one or more other dead and poorly attested languages of the Balkans?Footnote ¹⁴ Woodard 2004: 14 adduces the example of Ancient Macedonian kebalá, and an apparently related form gabalá glossed as ‘head’ by Hesychius without identifying the language, and Greek κεφαλή ‘head.’ This and other glosses (e.g., AMac abroutes ~ AGrk ὀφρύες ‘eyebrows,’ from PIE *H₃bhru-, cf. Skt bhrū-, Eng brow) suggest that Ancient Macedonian retained the voicing of Indo-European voiced aspirates (e.g., *bh), but, perhaps, lost the aspiration, while Greek (or Hellenic) kept the aspiration but not the voicing. The question is unduly complicated by modern nationalist attitudes, according to which a determination of the provenance of Ancient Macedonian would (somehow) justify modern-day territorial or ethnonymic claims. We can note here that Hall 1997: 62–64 makes it abundantly clear that regardless of the linguistic relation of Ancient Macedonian to Ancient Greek, the weight of evidence indicates that the Ancient Greeks did not consider the Ancient Macedonians to be ethnically Greek (see also Badian 1982), and in any case, the ancient territorial and ethnonymic situation is utterly irrelevant for the modern one (cf. also Ilievski 1988, 1997, 2008; Joseph 2024). Our knowledge of Ancient Macedonian comes in large part from glosses in Hesychius, a fifth-century CE glossator and antiquarian with an interest in lexical curiosities of earlier Greek and in compiling them. Various texts and dialect materials were available to him that no longer exist, and he had earlier lexicographical works to draw on too. He listed several words said to come from “Μακεδόνες,” i.e., the ancient Macedonians. In addition, there are several place names and personal names that are either known from historical sources to be Ancient Macedonian or are presumed to be so.

It is important to note that no verifiable clearly Ancient Macedonian inscriptions exist.Footnote ¹⁵ For details on the existing data for Ancient Macedonian and their analysis, see Ilievski 1988, 1997, 2008; Katičić 1976: 100–116; Pudić 1967; Woodard 2004: 12–14; Neroznak: 1978: 168–173; and Gindin 1987, as well as the various pro-Greek articles in Giannakis 2012, with their detailed bibliographic coverage. Evidence concerning the Epirotes is likewise scant and uncertain (Katičić 1976: 120–127).

1.2.1.2 Illyrian

It is a widely held opinion that Illyrian represents the ancestor of Albanian (cf. Katičić 1976: 184–188; Polomé 1982: 888; and Ismajli 2015 and sources cited therein). The existing data, however, are so sparse, contradictory, and/or speculative, that such a claim has been challenged. Hamp 1994a, 1994b went so far as to question whether the term Illyrian, even in the narrow sense, actually refers to a single language or was rather a cover term used by ancient authors for a group of languages spoken by various Indo-European tribes that may or may not have been more closely related. (Cf. popular usage of the term Australian Aborigine, e.g., in the BBC television series “Dr. Who” [“Four to Doomsday,” Episode 1, first aired January 18, 1982], as if it referred to a single language rather than hundreds of different languages.)Footnote ¹⁶ As Woodard 2004: 11 notes, we do not possess a single verifiable inscription in Illyrian. Most of the evidence for Illyrian that is adduced is based either on onomastics – which, in addition to being highly speculative (we cannot be sure that an element found in a proper name actually corresponds to a lexical item that it happens to resemble in some other language), clearly does not form a single consistent whole (Polomé 1982: 866; Sh. Demiraj 2004: 55–56; cf. also Katičić 1976: 154–166, Tzitzilis 2007a: 746) – or on the problematic assumption that Messapic, a poorly attested and inadequately deciphered Indo-European language of southern Italy, was a relative or dialect of Illyrian (see Polomé 1982: 866; Hamp 1994b).Footnote ¹⁷ As Woodard 2004: 15 observes, however, until and unless Illyrian is better attested, a connection with Messapic must remain an unverifiable hypothesis (but see now Hamp 2008).Footnote ¹⁸ Our total unambiguous set of attested lexical items that are labeled by ancient authors as Illyrian consists of three words – rinos ‘mist,’ sabaia ‘fermented fruit beverage,’ Deuadaia ‘satyr’ (Katičić 1976: 170–171; Tzitzilis 2007a: 746) – and a possible fourth sybina or sygine ‘hunting spear’ (Duridanov 1999: 755; Polomé 1982: 867).Footnote ¹⁹ The first of these is plausibly connected with Alb rê (older ren) ‘cloud,’ the second seems cognate with Eng sap (Skt sabar), the third may be connected to Grk θύω ‘rage,’ and, if the fourth is admitted, a cognate with Armenian suin [səvin] ‘spear’ has been suggested, but this is hardly adequate proof that we are dealing with a single language, much less one whose modern descendent can be determined with certainty.Footnote ²⁰ Other evidence connecting Illyrian with Albanian is similarly flawed. Thus, for example, the toponym Dalmatia has been associated with Alb dele ‘ewe,’ with variant form delmë (Meyer 1891: s.v.), supported by evidence from Strabo (7.5.5) referring to Delminium, the capital of the Delmatae, as πεδίον μηλόβοτον ‘a pasturage for sheep’ (cited in Sandfeld 1930: 143). In fact, however, the passage in Strabo does not say that Delminium was a pasturage for sheep, but rather that Nasica turned it into one, i.e., he razed it – the expression is an idiom for utter destruction and cannot be interpreted as a literal reference to sheep (Katičić 1976: 173). The resemblance of the ethno-toponym Dardania to Albanian dardhë ‘pear’ (cited in Sandfeld 1930: 143) is suggestive, but nothing more. As with the Ancient Macedonians, so too with the Illyrians, at times modern nationalist claims of legitimacy based on autochthony or illegitimacy based on migration interfere with objective evaluations of the available evidence; and as with the Ancient Macedonians, so too with the Illyrians, such considerations have no place in modern linguistics nor should they be taken as the basis for modern geopolitics. Regardless of what the ancient situation may have been, the modern one is what we have, and the question of Illyrian remains open, even if some sort of connection with Albanian seems reasonable. In any case, as with the other dead languages of the Balkans, we do not have a single shred of unequivocal textual evidence above the level of the lexical item.

1.2.1.3 Thracian

We have considerably more lexical evidence for Thracian than for either Ancient Macedonian or Illyrian, e.g., brûtos ‘beer’ (cf. OE breowan ‘to brew’ (Woodard 2004: 12)), and classical sources tell us that it was widely spoken in the region that we call the Balkans today. According to Strabo, the Getae shared their language with the Thracians and the Dacians spoke the same language as the Getae (Katičić 1976: 131–132). We know that Ovid wrote poetry in the Getic dialect, but none of it has survived (Katičić 1976: 137). We have two short inscriptions in Thracian, but, as Katičić 1976: 137 observes, they “are much too scant and their interpretation not sure enough to be of any real value for the study of Thracian.”Footnote ²¹ We also have what appears to be a Dacian inscription – Decebalus per Scorilo which Georgiev 1977: 199 glosses ‘Decebalus son of Scoril[o]’ (cf. Duridanov 1999: 749). However, as Brixhe 1994: 186 points out, the inscription is in running script and thus the segmentation is far from certain. Georgiev 1977: 193–215 has adduced evidence connecting Albanian with Thracian or Daco-Moesian, and as Fine 1983: 10–11 points out: “these are serious (nonchauvinistic) arguments and they cannot simply be dismissed.” Nonetheless, in the absence of more extensive, genuinely textual, evidence, we cannot be certain what, if any, the relationship was of Thracian (or Dacian/Daco-Moesian) to Albanian. Fine 1983: 11 further writes: “More evidence is needed which, owing to the nature of our sources, may never be obtained; thus the question may well be one of many in early Balkan history which we may never be able to answer.” The possibility of a relationship between Thracian and Phrygian is likewise uncertain and vexed (Woodard 2004: 12; Brixhe 2004: 780).

1.2.1.4 Dacian (Daco-Moesian)

DacianFootnote ²² is identified as the Indo-European language spoken on that part of the territory of modern Romania that was occupied by the Romans starting with the first battles in 101 CE to the organized migration south of the Danube in 271 CE. Moesian has been identified as a language or dialect south of the Danube closely allied with Dacian (Georgiev 1977: 193–211). While modern Romanian is a Romance language, it is generally assumed that remnants of the Dacian lexicon persist, and moreover the existence of a group of about seventy to ninety Indo-European cognates shared between Albanian and Romanian, e.g., mal Alb ‘mountain’ – Rmn ‘river bank,’ Alb sorrë – Rmn cioară ‘blackbird’ (from PIE *k^wērsnā ‘the black (thing)’), Alb moshë ‘age’ – Rmn moş ‘old man,’ is taken as indicative of a close prehistoric connection between the ancestor of Albanian and the pre-Latin language of the territory that is included in modern Romania.Footnote ²³ Given that Dacian (or Daco-Moesian) appears to have been closely allied to Thracian and that the evidence that is taken as constitutive of Illyrian appears to belong to a different Indo-European group, the question naturally arises whether the evidence of the Albanian–Romanian connection points to a genetic/genealogical or areal link. Hamp 1994a is quite clear in suggesting that Romanian and Albanian appear to have shared in intense Romanization, with the language that became Romanian having been more or less fully Romanized while that which became Albanian having been on the road to such a shift but not having undergone it. But, as noted above, Hamp is extremely cautious regarding nomenclature and suggestions of genetic versus areal relationship in this regard, and we share in his caution.Footnote ²⁴ Hamp 1994b, 1999 uses the term Albanoid to denote the stage of Albanian after it diverged from the Balto-Slavo-Albanian dialect of late Proto-Indo-European, but before it came into contact with Latin and the Roman empire (circa 200 BCE). Proto-Albanian refers to the period when the language was in contact with Latin, but before contact with the Slavs (circa the seventh century CE). Common Albanian refers to the variety of Albanian that all extant varieties must have descended from.

1.2.1.5 Pre-Greek, Pre-Hellenic, or “Pelasgian”

There are numerous words that are part of the Ancient Greek lexicon that are clearly not Hellenic, that is, not direct inheritances from Proto-Indo-European. While some of these are borrowings from identifiable sources, e.g., χιτών ‘tunic’ from a Semitic language (cf. Phoenician ktn ‘linen tunic’) or κύανος ‘copper’ from a language of Anatolia (cf. Hittite kuwanna(n) ‘copper’), many do not have a recognizable source. Moreover, some of these words are strikingly similar to – but, crucially, not identical with – existing inherited Greek words, e.g., ἀλείφ-ω ‘anoint,’ which in its -λειφ- “core” seems quite similar to the root of the inherited lexeme λίπ-ος ‘fat’ (thus, identical but for the initial vowel and final aspirate). Furthermore, place names all over Greece have recurring apparently suffixal elements added onto stems that have no clear Greek parallels, e.g., the -νθ- found in the toponyms Κόρινθος and Ζάκυνθος, among others, and there are some words for items of material or natural culture that have similar elements, such as ὑάκινθος ‘hyacinth’ or ἐρέβινθος ‘a kind of bean.’ Putting all this together, scholars have posited, quite plausibly, at least one lexical layer, and maybe more, of words that entered Greek from an indigenous, most likely Indo-European-speaking, group that populated Greece before the coming of the Greeks early in the second millennium BCE. The language or languages of this group are usually referred to as Pre-Greek (Katičić 1976: 16–97), or Prehellenic (Hamp 1994b), and sometimes also “Pelasgian” (using an ethnonym that occurs in the Odyssey and in other ancient sources), and it is possible that other groups are to be identified in other lexical layers. The toponyms, moreover, are generally held to reflect a language of ancient Anatolia, quite possibly (but by no means assuredly) an Indo-European one, given apparent parallels in the shape of certain suffixes (e.g., -anda- in Anatolian place names, such as Kuranda, cf. Hittite kuera- ‘field,’ reminiscent of Κόρινθος). As interesting as this material is (even though much of it is speculative and based on similarities of form that might be accidental), and as important as it is for understanding the prehistory of the Balkans and related matters such as the coming of the Greeks into Southeastern Europe, it is entirely lexical in nature and moreover none of the languages possibly identified in this way can be shown to have contributed anything to the pan-Balkan features that emerged centuries later and gave rise to the Balkan sprachbund.Footnote ²⁵

1.2.1.6 Phrygian

According to ancient sources Phrygian was spoken in the Balkans during the prehistoric period (Katičić 1976: 130–131; Tzitzilis 2007b; Brixhe 2004: 777). Although the language is poorly attested, we have whole inscriptions in it, but all of them are from Anatolia (Hamp 1994b), and even in these the grammatical evidence they give us is meager (Brixhe 2004: 782–786). As a Balkan language, therefore, Phrygian is without adequate concrete textual evidence.Footnote ²⁶

1.2.1.7 Paeonian and Dardanian

These are ethnonyms associated with territory approximately in the center and south of today’s Republic of North Macedonia and in Kosovo, respectively. The languages to which they presumably refer are so poorly attested and understood that their relationship to the other dead languages of the Balkans is a matter of numerous contradictory speculations linking them with Thracian, Dacian, and/or “Illyrian” (Solta 1980: 35–39; Sh. Demiraj 2004: 67; Katičić 1976: 117–118, 181; Duridanov 1999: 746). As Katičić 1976: 118 points out: “Paeonia is a blank on the linguistic map of the ancient Balkans.” The only gloss we have is mónapos ‘European bison’ (cf. Thracian bólinthos, Woodard 2004: 12). Papazoglu 1978: 218–219 notes that we have three glosses in Dardanian, all of them plant names: aloítis ‘gentiana,’ sōpîtis ‘wormwood,’ and kakalía of uncertain meaning. While some have connected sōpîtis with Illyrian sabaia, there is no way to test the association. Moreover, even if all three plant names were to prove to be Illyrian (and Illyrian an identifiable language), we would have no way of knowing whether or not they were loanwords. As noted in §1.2.1.2, the connection of the name Dardanian with Albanian dardhë ‘pear’ is suggestive but unsubstantiated. In the early 2000s, Kosovar Albanian nationalist practice has sought to replace the toponym Kosovo (a Slavic/Serbian denominal adjective from kos ‘blackbird’ understood as modifying the neuter noun polje ‘field’) with the resuscitated form Dardania.Footnote ²⁷ With independence in 2008, however, the internationally recognized name of the country is Kosovo.

1.2.1.8 Celtic

The Celts were very much a presence in Southeastern Europe in ancient times, having invaded the Balkans in 279 BCE. They established settlements thereafter in what is now central Serbia, Thrace, and the Danube basin in parts of present-day Romania and Bulgaria, and they raided as far south as Delphi. Little is known about the Balkan Celtic language(s) – there may well have been more than one Celtic tribe involved – beyond personal names (see Katičić 1976: 180ff.; see also Papazoglu 1978: 50–56, 340–389, 439–491 et passim, and Sandfeld 1930: 98). However, there are a few inscriptions from the general area of ancient Noricum in southern east central Europe in what could be classified as the northern extremes of the Balkans, depending on how “Balkan” is defined (see §1.1), in particular from Ptuj (Slovenia) and Grafenstein (southern Austria). Intriguingly, the Grafenstein inscription has a sequence of letters which read ollo=so.Footnote ²⁸ Some scholars have seen in this sequence ‘all’ (cf. Old Irish oll ‘ample’) followed by a postpositive demonstrative/article (cf. Sanskrit sa, and other Indo-European *s- demonstrative elements), perhaps as a noun phrase in itself or as a modifier of a subsequent word. Such a reading is strikingly reminiscent of the Balkan postposed definite article.Footnote ²⁹ Given that the inscription is essentially uninterpretable and that this “identification” rests entirely on etymological guesswork (however enlightened it may be), it is hard to make any legitimate connection between this sequence and the better understood Balkan phenomenon. Moreover, it is not clear that this is even the same language as that used by the Celts situated more in the central and southern Balkans. Thus however tantalizing this evidence may be, it does not offer much of substance to our understanding of the ancient Balkan linguistic situation or to later sprachbund phenomena.Footnote ³⁰

1.2.1.9 Germanic

Germanic tribes such as the Gepids and Goths also settled in the Balkans during the late ancient period. Moreover we know that the Norse Vikings were in the Balkans since there are Runic Norse inscriptions in a few places, as described by Page 1987: 53, most notably a grafitto of “the name Hálfdanr scratched on the marble of the great church of Hagia Sophia, Istanbul, and an inscription, hardly readable these days, cut on a marble lion once at Piraeus and now outside the Arsenal, Venice.” Nonetheless, their languages do not appear to have had any discernible effect, beyond toponyms and perhaps some other vocabulary, mostly for items of material culture (Mihaescu 1993: 322–323; cf. also Poruciuc 2009), but also ecclesiastical terminology that entered OCS via Frankish or OHG (Lunt 1982). Also, Jokl 1929 identified a few cases of loanwords in Albanian specifically from Gothic, such as fang ‘sod, piece of turf’ from Gothic waggs ‘Paradise’ and he suggested further that tirk ‘trousers’ is to be connected with a presumed Gothic *þiu(h)-brók-, (lit., ‘thigh-breeches’), a form safely inferrable for Gothic due to attested OHG theobroch (see also Barić 1954). Albanian tirk is of particular interest since it is related to Romanian tureac ‘leggings’ (with a by-form tureatcă), which is analyzed by Diculescu 1929 as also being based ultimately on this Germanic form (though he sees it as transmitted through Gepids). Although Sandfeld 1930: 97 raises the question of whether Low Latin tubroces could be an intermediary in the importation of this Germanic form, at which point the Romanian and Albanian words would not be direct evidence of Germanic influence in the Balkans, the analysis by Hamp 1990 has clarified how both the Romanian and the Albanian forms represent different developments out of a Germanic, and specifically Gothic, starting point.

1.2.1.10 Iranian

Iranian-speaking tribes such as the Scythians and Sarmatians extended into the Balkans and left lexical traces, e.g., in hydronyms such as Danube. Iranian military aristocracies are also associated with the Slavic invasions of the early Middle Ages (Fine 1983: 26, see §1.2.3.4). There are no Balkan Iranian texts, however, and so Iranian joins Germanic and Celtic as a transient linguistic group in the Balkans.

1.2.1.11 Pre-Ottoman and Post-Ottoman Turkic

A number of different Turkic-speaking tribes passed through or settled in the Balkans prior to the arrival of the Ottoman Turks (cf. Girfanova 2001), e.g., Pechenegs, Kumans (Polovtsians), Huns, Avars, and Bulgars. Some of these groups were Kipchak (Northwest Turkic), while others (e.g., Huns and Bulgars), apparently, belonged to a branch whose only living descendant is Chuvash (the so-calledl/r branch as opposed to the s/z branch, to which all other living Turkic languages belong, e.g., Chuvash tăxăr = Turkish dokuz ‘nine’). The Bulgar Turks (sometimes called Proto-Bulgars), who were assimilated by a group of Slavs who took on their ethnonym (the Bulgarians), have left lexical traces in the Balkans (e.g., Sandfeld 1930: 99; Parzymies 1994; Dybo 2012) but, based on the evidence of Old Church Slavonic, no discernible morphosyntactic effects (but cf. Johanson 1996, 2001; Mladenova 2007: 357–362). It has sometimes been claimed that Gagauz, which is generally classed as Oghuz (Southwest) Turkic (as is Turkish), contains Kipchak elements, suggesting possible survivals from pre-Ottoman Turkic peoples (e.g., Kumans, cf. Johanson 1992, 1996, 2002), but these claims are not considered firmly established. It is now generally accepted among historians that the Gagauz are descended from Turkish mercenaries from Anatolia settled in Dobrudja by the Byzantine emperor Michael VIII in the mid-thirteenth century (Wittek 1953; Fine 1987: 215, but cf. also Pokrovskaja 2001). Mollova 1966: 139 notes that Gagauz speakers in Bulgaria consider their language to be a Turkish dialect and observes that speakers refer to their language as bizim türkçemiz ‘our Turkish.’Footnote ³¹ Finally, we can note here that a significant number of Tatars settled in Dobrudja and many entered Bulgaria in the wake of the Crimean War (1854–1856). Although small groups of Tatars still live in the Balkans today, Tatar has not participated significantly in the long-term processes under consideration here.

1.2.2.1 Hungarian

Spoken throughout the former Hungarian Crown lands, Hungarian is sometimes mentioned in studies dealing with the Balkan languages. For example, Schaller 1975: 151, citing Večerka 1970: 246, connects the Hungarian formation of teens with Slavic influence, e.g., tisz-en+egy ‘ten-superessive+one = one-on-ten = eleven’ (see §6.1.6), but he neglects to mention that the twenties are also formed on that model, e.g., husz-on+egy ‘twenty-superessive+one = twenty-one,’ which could not be a calque on Slavic.Footnote ³² We can also note here that Hungarian has a special treatment of definite direct objects in that it marks them on the verb in its so-called transitive conjugation.Footnote ³³ Masica 2001: 243–245 links this with Romanian definite direct object marking by means of the preposition pe, but he neglects to include the resumptive pronominal clitics (on which see §7.5.1) of Balkan Romance, Balkan Slavic, Albanian, and Greek (and to a lesser extent, Romani), which then connects the region with Turkish and beyond even to South Asia. Hungarian also has definite and indefinite articles, but they are both preposed, as in German (§6.1.2.2.1). In any case, the status of Hungarian with respect to the Balkan sprachbund is basically peripheral, a source of some vocabulary, e.g., Slv varoš ‘old town,’ Rmn oraş ‘town, town center,’ Alb varosh ‘suburb, neighborhood,’ Grk βαρόσι (also βαρώσι) ‘suburb, quarter,’ Trk varoş, all from Hungarian város (Sandfeld 1930: 98; Schubert 1999).

1.2.2.2 German

German was spoken in former Habsburg lands and also in colonies of workers brought to Ottoman lands for their technical skills but expelled en masse from Yugoslavia after World War Two. Some speakers also emigrated or were expelled from other Eastern European countries. A significant, albeit dwindling, minority continues to live in Romania.

1.2.2.3 Armenian

Armenian was first brought to the Balkans by the Byzantines in the ninth century, with communities coming also from subsequent migrations. The most recent migrations arrived in the wake of the Armenian genocide in Anatolia at the beginning of the twentieth century; there are still Armenian communities in North Macedonia, Bulgaria, and Greece. Their dialects have not received adequate attention, but Adamou 2008 notes that the Armenian of Thessaloniki is considered distinctive in its avoidance of the infinitive.

1.2.2.4 Georgian

A Georgian monastery (P’et’ric’on) was founded at Bačkovo in the Bulgarian Rhodopes in the eleventh century (construction completed 1083), when the Georgian province of Tao and Bulgaria were both part of the Byzantine Empire. Two inscriptions and old copies (thirteenth century) of the monastery’s tipikon survive (Šanidze 1971).

1.2.2.5 Circassian (Adyghe)

A significant number of Adyghe Circassians were settled in the Balkans by the Ottomans as a result of migrations caused by the Russian conquest of the Caucasus (1769–1864). According to Kurmel 1994 the language was retained in at least one village in Kosovo about 10 km from Prishtina, Čerkes Kjoj (probably the same as current Balshaj (Alb)/Miloševo (Srb)), until the end of the twentieth century, but according to Kumpilova 2012 there were still several Circassian villages in Kosovo as well as some Circassians in Gjilan/Gnjilane up to the 1998/1999 war, when most of those remaining were evacuated to Adyghea. According to Sikimić 2005c, however, there are still some Circassians in Balshaj/Miloševo and in Prishtina. Kânčov 1900: 116, 178, 215 gives locations and statistics for Circassians in Macedonia, but their dialect is completely lost to us (for Kosovo, see also Vermić 2005).Footnote ³⁴

1.2.2.6 Yiddish

Yiddish was spoken in the Austro-Hungarian Balkan lands and by Jewish immigrants from Austria-Hungary, Germany, and the Russian Empire as those states’ interests penetrated the southern Balkans. Significant numbers of Yiddish-speaking Jews settled in Wallachia and Moldavia considerably earlier. Ashkenazic Jews were also present in the Balkans from earlier periods. Most were annihilated by the Nazis in World War Two, and most survivors emigrated to Israel, but some communities remain.

1.2.2.7 Other Languages in the Balkans

In addition to these languages, we do not treat the following languages that belong to language groups that are otherwise represented in the Balkan sprachbund.

1.2.3.1 Albanic

There is no doubt that Albanian is descended from one of the poorly attested or unattested ancient languages of the Balkans, but, as is clear from the discussion in §1.2.1, the exact identification cannot be made with any certainty (but see §1.2.1.2, Footnote footnote 18). Hamp 1994a argues on the basis of shared innovations such as the Winter’s Law vowel lengthening (see Footnote footnote 18) that within Indo-European, Albanic was closest to Balto-Slavic (see Hyllested & Joseph 2022 on Albanic forming a subgroup with Greek) and this shared innovation together with shared ancient non-Indo-European borrowings such as the word that became ‘apple’ in Germanic and Balto-Slavic but ‘sorbus’ in Albanian – vadhë (Geg vodhë, vollëz, etc.) – indicates that during the late Indo-European period, the dialect that ultimately became Albanian was part of the North (Central) European area (Hamp 2010). At some later point, Albanoid, i.e., the Indo-European language that became Albanian prior to its contact period with Latin, was spread along the Carpathians from southeastern Poland to Croatia, as attested by such widespread lexical items as vatra ‘hearth’ (Hamp 1976, 1981), strungă (Rmn)/shtrungë (Alb) ‘enclosure for milking animals’ (Hamp 1977b). Hamp 1999 uses the term Proto-Albanian for the period of contact with Latin and describes this period in the following terms: “The eastern portion of this speech area adopted Latin but kept traces of its old grammar and many lexemes. The result is called Romanian. […] The western portion accepted many loans but kept its language” and became Albanian (Hamp 1994a: 67). In Hamp’s terminology, Common Albanian refers to the period after the contact with Slavic but before the dialectal split between Geg and Tosk. Attested Albanian begins with a baptismal formula embedded in a Latin text from 1462, and the first major dated textual attestation is the 1555 Meshari ‘Missal’ of Gjon Buzuku (Çabej 1968).Footnote ⁴²

The Geg/Tosk split is defined by a relatively compact bundle of isoglosses whose northern border (rhotacism and the development of *uo into ue (Geg) or ua (Tosk) in syllables closed by a sonorant) runs along the course of the river Shkumbî in central Albania and whose southern edge (denasalization and the development of stressed schwa) is roughly 10–20 kms to the south (with the change of *vo- > va- located in between).Footnote ⁴³ It represents a defining moment in the history of Albanian unlike any other since contact with Latin. Although there are other features participating in the split (e.g., infinitive with me, sigmatic imperfects), the key phonological features are both emblematic and, in a certain measure, identifiable in temporal terms. That the split post-dates contact with Latin is certain, since Latin loanwords uniformly undergo the changes that characterize the split, e.g., orphanus > vorfën/varfër ‘poor,’ (h)arēnus > rânë/rërë ‘sand.’ Equally certain is that the changes had occurred before the diasporas of the late medieval and early modern periods that produced Arvanitika and Arbëresh, which are unmistakably Tosk (see below). Between the terminus a quo and the terminus ad quem, however, stretches a period of seven or eight centuries during which another event significant for the history of Albanian occurred: the arrival of the Slavs.

There are three main proposals concerning the chronology of the Geg/Tosk split: (1) It was completed before contact with Slavic (Gjinari 1989: 2); (2) It began prior to but continued into the period of contact with Slavic (Hamp 1994a; Rusakov 2013); and (3) It began after contact with Slavic (Janson 1986). Without entering here into the details of the arguments adduced, we note simply that the evidence of Slavic loanwords for the chronology of the Geg/Tosk split is sufficiently meager and complex that any given account must explain away counterexamples. On the one hand, Slavic loans in /n/ that do not rhotacize in Tosk can be explained as analogical or late (e.g., suffixal -nik, stopan ‘herdsman’) while on the other, rhotacized examples of obvious antiquity such as Geg shtëpâ/shpneshë ‘herdsman/housewife,’ the latter equivalent to Tosk shtëpreshë ‘housewife,’ from Slavic stopan ‘master, householder,’ can only be explained away by claiming a non-Slavic intermediary (e.g., the ancestor of Romanian stăpîn ‘master, owner’) or by arguing that rhotacism continued as a local phenomenon in isolated areas. If, with Hamp 1994a: 67 and Jokl 1923: 42, 46, we accept the treatment of Slavic *č as *s in Albanian porosit ‘order’ from Slavic porǫčit- as evidence that initial contact with Slavic took place prior to or during the Common Albanian period, then the issue of the exact dating of rhotacism, however important it may be for Albanian etymology, becomes moot for the question of Slavic-Albanian language contact, since by definition Common Albanian pre-dates the Geg/Tosk split (see also Rusakov 2013 on the relative dating of Albanian isogloss bundles.)

Although other, more recent, isoglosses sometimes traverse the Shkumbî (Maynard 2002; Friedman 2003f), Albanian dialectological divisions are basically subordinate to the Geg/Tosk distinction. Within Geg we can differentiate Southern Geg, East Central and West Central Geg, Northeast Geg, and Northwest Geg. Southern Geg stretches from the Adriatic to just beyond Mts. Jablanica and Belica in North Macedonia and includes Elbasan and Tirana. This dialect functioned as the basis of a de jure, but not always de facto, Albanian standard from 1923 until the early 1950s. West Central Geg is entirely in Albania, north of Southern Geg, and East Central Geg begins in Albania just west of Mts. Korab and Jablanica and then extends into North Macedonia, accounting for the great majority of Albanian dialects there from Tetovo and Kumanovo almost to Bitola.Footnote ⁴⁴ The dialects of Montenegro and Shkodër are Northwest Geg, while Kosovo and adjacent parts of Albania and some Macedonian border villages speak Northeast Geg. The only older Geg diaspora dialect is that spoken in the village of Arbanasi (Itl Borgo Erizzo), a suburb of Zadar, to which inhabitants of two villages near Bar (Itl Tivari) emigrated in the eighteenth century. There is also evidence of a Geg dialect in Istria (Altimari 2011a) and Srem (Hamp 1994a). The Northern Tosk dialects occupy most of southern Albania and extend into the southwestern corner of the Republic of North Macedonia and much of the western part of northern Greece. They also serve as the basis for modern Standard Albanian (Byron 1976). The Lab dialects are spoken south and west of the Vjosa River and extend into Greece between Northern Tosk and Çam. The Çam dialects begin between Butrint and Konispol in southern Albania and extend into Epirus/Çamëri. Although exempted from the exchange of populations between Greece and Turkey as arranged in the Treaty of Lausanne in 1923 (despite the fact that they were Muslims),Footnote ⁴⁵ most Çams were expelled from or left Greece in the wake of the Civil War in 1948, although there are still remnants of the Albanian-speaking population in Epirus (Rexhep Ismajli, p.c.).

There are three main Tosk diaspora dialects. Arvanitika separated from southern Çam by the thirteenth century or earlier and was spoken all over southern Greece and on many of the islands into the twentieth century (see Hamp 1961). The language, which its speakers call Arbërisht, has undergone massive attrition under pressure from Greek since World War Two and has disappeared completely in places where in the 1950s there were still significant numbers of speakers (Hamp, p.c., see also Tsitsipis 1998). Arbëresh, still spoken in forty-five to fifty villages of southern Italy (Altimari 1993; Hamp 1994a; Nasse 1964), separated from Arvanitika and Albanian under pressure from the Turkish conquest, especially after the death of Skanderbeg in 1468 and after the reprisals for the rebellion of 1481. Another exodus to Italy occurred in 1492 (Fine 1987: 602). The one other set of migrations (which probably involved more than one wave) was eastward mostly from the northern Tosk region, but with some southern Tosk elements (Hamp 1965; Liosis 2021) also in the early modern period, to what is today Bulgarian, Greek, and Turkish Thrace and to northeastern Bulgaria, and some speakers continued from northeastern Bulgaria to Southern Dobrudja (Budžak) and Crimea in the nineteenth century. Some Albanian speakers, who found themselves on the Turkish side of the border after the partition of Thrace, migrated across the border to the Greek side (Friedman 2004a: 59–155; Novik et al. 2016: 39–358; Dalatsis 2016; Kotova 2017; Johalas 2019; Liosis 2021). These speakers, like the Albanian speakers of Albania and contiguous areas, generally call their language shqip, which is probably related to shqiptoj ‘pronounce clearly’, cf. Latin explicare (Hamp 1994a). However, some villages simply use si neve ‘like us’ (cf. the Macedonian speakers of Boboshtica and Drenova (Boboščica and Drenov^jäne, in the local dialect), in Albania who use kaj nas ‘like us’). The villages of Léhovon (Mac Lehovo), Drosopigi (Mac Bel Kamen), Flámbouro (Mac Negovani), and Eláteia (Mac Elovo), south and east of Flórina (Mac Lerin) in what is now Greek Macedonia, were settled by Christian Albanian- and Aromanian-speakers from Epirus at the beginning of the nineteenth century (Simovski 1998[2]: 137–182). At one time there were also Albanian villages across the Rhodope range and into Thrace, then up the Black Sea Coast to the Danube. Of these, a single village in Bulgaria, Mandrica, near the Turkish and Greek borders, as well as some neighboring villages on the Greek and Turkish sides, survive (Hamp 1972b; Sokolova 1983; Stankov 2016; Liosis 2021). An offshoot of Mandrica from the Balkan Wars is Mandres, south of Kilkis (Kukuš) in Greek Macedonia (Hamp 1965). Three villages of Albanians were also settled near the Sea of Azov in Ukraine in 1862, after the Crimean War (Kotova 1956: 254–255). See Friedman 1994c for a discussion of the dialects of Turkish Thrace and Liosis 2021 for Greek Thrace. Many Albanian-speakers left what is now former Yugoslavia for what is now Turkey as the Ottoman Empire contracted, especially in 1878–1923, and many more left Yugoslavia in the 1950s. Their dialects survive to varying degrees.Footnote ⁴⁶ See also Desnickaja 1968 on Albanian dialects.

1.2.3.2 Hellenic

The Greeks came to the Balkans during the first half of the second millennium BCE, displacing or absorbing other Indo-European and/or non-Indo-European peoples (Browning 1983: 1; Drews 1988; Gindin 1967; Horrocks 2010: 9).Footnote ⁴⁷ Although Greek was first written using the Mycenaean syllabary (so-called Linear B) in the second half of the second millennium BCE, the syllabary was lost with the collapse of Mycenaean civilization (c.1200 BCE) and the Greeks reverted to illiteracy. Eventually the Phoenician alphabet was adapted to the representation of Greek, and by the eighth century BCE inscriptions in Greek are to be found and there are as well works that can be called literary (Browning 1983: 3). From then on into the Hellenistic and Roman periods (c.300 BCE to 300 CE), there was a flowering of production of literary, historical, and scientific works in Greek, and literally thousands of inscriptions of all sorts (official decrees, treaties, dedicatory and funerary epigraphs, informal graffiti, and so on) were produced. By the first century CE, however, as the result of influence from (and reverence for) classical models of Greek usage, written Greek had become sufficiently divorced from the spoken language that for more than a millennium much of the textual evidence is never better than equivocal in terms of representing contemporary usage. Nonliterary papyri, however, containing personal letters, among other types of documents, do provide some insights into the language as found in more mundane day-to-day uses up to approximately the seventh century CE. In the twelfth century we begin to find texts such as the poems of Ptochoprodromos and Michael Glykas that do not attempt to imitate in a puristic fashion earlier literary (generally Attic Greek) models and can be taken as more reliable representations of what the spoken language may have been like. This is generally viewed as the beginning of Modern Greek. Nonetheless, even these late medieval and early modern texts have features whose status as puristic versus vernacular cannot be established with certainty. The Ερωτόκριτος of Vintsentzos Kornaros of seventeenth-century Crete represents the culmination of this tradition (Browning 1983: 7–9). The competition between the consciously archaizing and Atticizing Katharevousa (Puristic) Greek and the vernacular-based Dimotiki (Demotic) Greek, begun in the nineteenth century, has been decided in favor of Dimotiki, though the influence of Katharevousa on Dimotiki is undeniable, mainly in matters of vocabulary, phonology, and morphology. Nonetheless, Katharevousa is the language of an important body of literature and continues to interact with Dimotiki (on which, see Mackridge 1985, 1990; Kazazis 1992, among others).

Although Ancient Greek was divided into a number of dialects, the Koine (‘common’) dialect, based on Athenian Attic with significant input as well from the Ionic dialect, had displaced all of them by the end of the second century CE (Browning 1983: 50–52). The only substantial remnant of any other Ancient Greek dialect is Tsakonian, which is descended more or less directly from Ancient Doric (the Laconian variety) and spoken in the mountains of the southeastern coast of the Peloponnese and, until the exchange of populations between Greece and Turkey, on the southern shores of the Sea of Marmara by colonists from the eastern Peloponnese who had migrated there in the fifteenth century (Browning 1983: 124).Footnote ⁴⁸ There are also some ancient Doric elements in the Greek of southern Italy . All remaining dialects of Greek are descended from the Koine. On mainland Greece, another wave of dialect extinction began with the end of Turkish rule in the nineteenth century. Since the Peloponnese constituted the overwhelming majority of the territory of the early Greek state (and most of Attica and Central Greece was heavily Arvanitika-speaking), immigrants from the Peloponnese poured into Athens, the new capital, and the town dialect of Athens together with the dialects of Megara, Aegina, and Kyme in Euboea (the so-called Old Athenian dialect) which, among other features, had /u/ from Ancient Greek /ü/ (upsilon), was submerged in a flood of Peloponnesian.

The Peloponnesian dialects are relatively homogeneous, and the dialects of Istanbul and the Ionian Islands are relatively close to Peloponnesian (Mackridge 1985: 5). Of the central or core dialects, this leaves only the dialects of northern Greece as distinct.Footnote ⁴⁹ According to Browning 1983: 120–121, the main isogloss separating the northern dialects goes along the coast of Epirus and Acarnania, along the Gulf of Corinth and across the Isthmus, along the northern mountain frontier of Attica, south of Euboea, across the middle of the island of Andros, north of Icaria and south of Samos (but excluding Chios) and to the coast of Asia Minor. North of this line, unaccented high vowels (/i/ and /u/) are dropped (sometimes with palatalization of a consonant before unaccented /i/) and unaccented mid-vowels (/e/ and /o/) raise to high vowels. Northern dialects also have velar [l] before back vowels, pronounce /s/ as [š] before front vowels, and use the accusative, in some instances preceded by από ‘from,’ in place of various uses of the genitive.Footnote ⁵⁰ It is worth noting that the north is where settled populations speaking Albanian, Aromanian, Macedonian, Bulgarian, and Turkish (as well as Romani, Judezmo, and Meglenoromanian) were often in the majority, or plurality, depending on the region or village, and these distinctive features of northern Greek, if not contact-induced, are at the very least consistent with the larger linguistic environment, e.g., some of these same phonological processes occur in Albanian, Aromanian, and Balkan Slavic (in this region).

The exchange of populations with Turkey in 1923 resulted in the expulsion from Greece of 500,000 Muslims speaking all the languages of the Balkans, mostly from northern Greece, and the forced resettlement of 1,500,000 Orthodox Christians, many of them Greek-speaking but also speakers of Turkish, Romani, etc. The Christians from towns like Smyrna (Trk Izmir) were settled in the outskirts of Athens, while the Christians from rural areas such as Cappadocia and the Pontos (Black Sea) region, who were more likely to speak only Turkish, were settled in Aegean (Greek) Macedonia and other parts of northern Greece (see Ladas 1932; Pentzopoulos 1962). The new immigrants and their children assimilated to standard Greek although the older generations in Greek Macedonia also learned Macedonian.Footnote ⁵¹

The dialects of Mainland and Aegean Greece form a center in relation to which the remaining dialects constitute a periphery of varying degrees of antiquity and differentiation. Starting from the west, there are two enclaves in southern Italy , one each in Apulia and Calabria, which continue the dialect of Magna Graecia of the late Roman Empire (Browning 1983: 132), Crete, the Cyclades, the Dodecanese, Cyprus, Asia Minor, Ukraine (Rostov region and Mariupol in the Azov region, whose speakers came from Crimea), and the Caucasus (Abkhazia and southern Georgia). The Asia Minor dialects consisted of Bithynian in the northwest and Pontic in the northeast, related to dialects near Rostov in Ukraine, Livíssi (Trk Kayaköy) in the southwest, and Cappadocia, Phárasa (Trk Çamlıca), and Silli (Trk Sille) in central Anatolia.Footnote ⁵² Most speakers of Anatolian Greek dialects were resettled in Greece during the exchange of populations, and except for a few Pontic-speaking Muslim villages in the northeast, these dialects have disappeared from Turkey.Footnote ⁵³ Cappadocian, however, once thought to have died, is still alive and in use in villages in northern Greece (Janse 2009). Most speakers of Greek in the Caucasus left for Greece after the collapse of the Soviet Union and subsequent war in Abkhazia. The dialects of Mariupol and Anatolia show, as might be expected, influence from Tatar and Turkish, respectively. Thus, for example, Cappadocian Greek has the same eight-member vowel system as Turkish, while Mariupol Greek has lost the genitive and expresses possession by means of an izafet construction, e.g., σπίτι-τ πόρτα ‘door of house,’ τάτα-του σπιτ ‘his father’s house’ (Browning 1983: 135–136). As with the diaspora dialects of other languages, for the most part we do not consider these dialects here.Footnote ⁵⁴

Two Greek ethnolects that should be mentioned are those of the Sarakatsans, transhumant shepherds (now mostly settled) living in northern Greece, Bulgaria (where they are known as Karakačani), and the Republic of North Macedonia (where they are known as Sarakačani) and the Romaniote Jews, whose ancestors were living in Greece prior to the arrival of the Sephardim of the Iberian peninsula (see §1.2.3.3). Owing to their lifestyle and material culture, there had been speculation that the Sarakatsans were Hellenized Vlahs (see §1.2.3.3). As Høeg 1925–1926 demonstrated, however, the dialect of the Sarakatsans of Northern Greece is a straightforward northern Greek dialect that gives no indication of any unusual interference from Aromanian. Tzitzilis 1999 confirms that the dialect of the Karakačans of Bulgaria is of the same type. Most of the Sarakačans of what is now North Macedonia left for Greece during the 1960s owing to attempts at collectivizing their flocks (Nedelkov 2011).Footnote ⁵⁵

The Greek dialect of the Romaniote Jews is known as Yavanic (Yevanic) or Judeo-Greek. For the most part, Yavanic was supplanted by Judezmo after the arrival of a large number of Sephardim from Spain in 1492 and Portugal in 1496–1497, but some communities preserved the Romaniote liturgy, and Greek-speaking Jews produced written texts in Constantinople in the sixteenth century (see Hesseling 1897 for an edition of a 1547 translation of the Pentateuch into Greek, prepared for didactic purposes to help Jewish Greeks learn Biblical Hebrew).Footnote ⁵⁶ By the beginning of the twentieth century, Jews in the towns in Epirus of Ioánnina, Árta, and Préveza, and in Chalkída still spoke a form of Greek that differed in some features from the Greek of their Christian neighbors. The differences seem to be limited to phonetic, intonational, and lexical phenomena. In contrast to some other Jewish languages, no awareness of language separateness seems to have existed. As with many other Jewish communities with distinctive languages or ethnolects in Europe, the Holocaust destroyed the majority, most of the survivors went to Israel, and the language is in a state of attrition. See Connerty 2003 and Krivoruchko 2011 for studies of Yavanic.

One problem with the study of Modern Greek dialects is the relative paucity of material, especially in the more generally accessible languages. Moreover, for the most part, the most important primary descriptive and analytic sources treat island or diaspora dialects and thus are less useful with regard to Balkan Greek. In fact, as Kontosopoulos 1981: 131–132 points out, the dialects of much of mainland Greece, including Macedonia and Thrace, still have not been adequately described. Some notable exceptions are Mirambel 1929 and Pernot 1934, though they deal with the Peloponnesos, and useful summative works include Thumb 1912, Dawkins 1940, and most importantly Newton 1972 (with extensive reference to the relevant literature in Greek); Trudgill 2003 is an important reassessment of the dialect divisions of Modern Greek, and Tzitzilis 2022 offers a useful handbook survey (in Greek) of each of the major dialects.

1.2.3.3 Balkan RomanceFootnote ⁵⁷

Prior to Augustus (27 BCE–14 CE), Roman influence in the Balkans was limited to the coast, particularly the coastal towns. From the second century BCE through the first century CE, the Romans gradually annexed most of what is now the Balkans south of the Danube, and the second century CE saw their relatively brief (107–271) occupation of Dacia (roughly, modern Transylvania and western Wallachia). Studies of the language of inscriptions indicate that Roman linguistic influence extended southward to the so-called Jireček line (Jireček 1911). Petar Skok identified a somewhat different boundary, and the space between the two is now presumed to have been a zone of Latin-Greek bilingualism (Rosetti 1964a: 34–36).Footnote ⁵⁸ According to Jireček, the line began at Lezha (ancient Lissus) while Skok identified the southern limit of Latinity as beginning at Vlora (near ancient Apollonia). Jireček’s line then went east across Albania and North Macedonia, between Skopje (Scupi) and Stobi (near Prilep) south of Niš (Naissus) and Pirot (Turres) but north of Sofia (Serdica) and then across the Balkan (Haemus) range to Varna (Odessos). Skok’s line went northeast south of Ohrid (Lychnidus) and Skopje (Scupi) and north of Sofia (Serdica) to cross the Balkan range to Varna (Odessos). The Black Sea coast, however, was dominated by Greek as far north as Tulcea (Aegyssus) in today’s Romania (cf. Rosetti 1964a: 34–36, 1973: 47–48; Kaimio 1979: 86–89). South of these lines, the dominant language was Greek. East Balkan Romance in its various forms (Romanian, Aromanian, Meglenoromanian, and Istro-Romanian) is descended from the language of Roman colonists and Romanized peoples east of Dalmatia and, in all likelihood, both north and south of the Jireček line, since the Romans maintained garrisons to guard roads throughout the Empire. A contested issue between the Romanians and Hungarians, however, is the question of whether the Romanian of present-day Romania, especially Transylvania, is descended from the language of Romanized Dacians and colonists who remained behind after Trajan’s evacuation of 271 CE (and is thus ‘autochthonous’), or whether it is descended from the language of Romans and Romanized peoples living south of the Danube who did not cross over the Danube (or at least into Transylvania) until after the arrival of the Magyars in the late ninth century (cf. Fine 1983: 10; Saramandu 2003–2004 argues for Romance continuity in Romania, while Du Nay 1996 argues against it).Footnote ⁵⁹ Contributing to this problem is the enormous gap between the last Roman inscription (late sixth century CE, Minkova 2000) and the first dated surviving document in Romanian, the Letter of Neacşu of Câmpulung from 1521.Footnote ⁶⁰ As a result, the arguments on both sides are based on conjecture and circumstantial evidence of uncertain quality.

Among Romanian linguists, there is a disagreement between those who recognize Aromanian as a separate Balkan Romance language and those who would make of it (along with all of Balkan Romance) a dialect of Romanian despite the many differences and the fact that the two have been separated for about a thousand years.Footnote ⁶¹ Aromanian (also known as Vlah, see below) is spoken south of the Danube in modern-day southern Albania, northern Greece, North Macedonia, and adjacent parts of Bulgaria as well as by émigré colonies in Dobrudja and elsewhere. See Ivănescu 1980: 30–46 for a summary of the debate; cf. also Savić 1987; Bacou 1989; Peyfuss 1994; and Jašar-Nasteva 1997. Aromanian is recognized and used as a distinct language in the Republic of North Macedonia.

We should also note here the existence of Meglenoromanian, surviving in a handful of villages in the southeast of the Republic of North Macedonia and adjacent parts of Greece as well as among migrants in some of the towns.Footnote ⁶² At the beginning of the twentieth century, Meglenoromanian was spoken in about a dozen villages in what was then the Ottoman kaza (county) of Gevgelija, nahiye (township) of Karadžova, on territory that was divided between Greece and Serbia (eventually the Republic of North Macedonia) in 1913 (Atanasov 1990: 1–14; one village in the region, Livãdzi, was Aromanian-speaking, its inhabitants having arrived in the eighteenth century (Puşcariu 1976: 224)). The largest Meglenoromanian village, Nănti (Mac Noti, Grk Nótia) was Muslim and ended up within Greece’s borders. With the exception of a single family that converted to Christianity, of whom only a single member was still alive in 1984 (Atanasov 1984: 479), almost the entire village was sent to Turkey during the exchange of populations in the 1920s and settled in various parts of eastern Thrace and western Anatolia (see Kahl 2006).

Linguistically, Meglenoromanian is heavily Slavicized (e.g., it has borrowed the prefixal system of Slavic aktionsart), showing evidence of long and intense contact with Macedonian. This is in contrast to Aromanian, which, in its various dialects, shows significant influence from contact with Greek and Albanian as well as Macedonian. The most important historical linguistic question raised by Meglenoromanian, however, is whether it represents the language of a population that became linguistically separated from Common Balkan Romance at the same time as Aromanian or at a later date. Although Atanasov 1999 argues that Meglenoromanian represents a later break-off from Common Balkan Romance that arrived via the Morava and Vardar valleys rather than via the Rhodopes (this latter route was posited by Capidan 1943: 16–17), his chief arguments rest on shared archaisms (e.g., a preserved infinitive) rather than shared innovations, and such shared innovations as are cited could represent later parallel developments.Footnote ⁶³ In general, shared innovations link Meglenoromanian to Aromanian, e.g., the change of velars to dentals before front vowels (Todoran 1977: 102–109), although the separation probably occurred at an early date (see also Kahl 2006).

The term Vlah entered the Balkans via a Gothic (and therefore Germanic) intermediary which had it from a Celtic tribal name (Skok 1973: 606–609). It is recorded by Caesar as Volcae, by Strabo and Ptolemy as Ouólkai, and it was in the transfer to Gothic (as *walhs) via Latin that the ethnonym took on the meaning ‘foreigner’ or ‘those folks over there’ or ‘Romance speaker’ (and, later, also ‘transhumant shepherd’ and other meanings). The metathesis of Wal- to Vla- is typically South Slavic. In Greece, the use of Βλάχος to mean ‘shepherd’ is a transference of the ethnonym based on a profession or lifestyle commonly associated with an ethnic group. In Albanian, the opposite occurs, and çoban ‘shepherd’ comes to mean ‘Vlah.’ In Serbia and Bulgaria, the ethnonym Vlah is used to refer both to people from Wallachia (i.e., Romania south of the Carpathians) – and, by extension, Romania as a whole – and to Romance-speakers south of the Balkan range.Footnote ⁶⁴ Moreover, in Serbia the term Vlah is also used to refer to Romanian speakers in eastern Serbia around Negotin and the Timok valley.Footnote ⁶⁵ In this latter sense, it is an ethnographic term referring to a group that differed from Romanians not in language but in a specific set of historical circumstances that led to their settling in eastern Serbia during the Ottoman period. Former Yugoslav census figures classified these Romanian speakers as Vlahs together with the Aromanian-speaking Vlahs of North Macedonia, and so one must examine figures at the republic level for an accurate picture. While Aromanians themselves use the ethnonym Armîn (in the south) or related forms such as Rămăn (in the north), all etymologically from Romanus ‘Roman’ and historically involving loss of the short, unstressed /o/ and an elimination of the resulting /rm/ as an initial cluster, Meglenoromanians designate themselves with the Macedonian form Vla (plural Vlaš) in their own language. (They are also known in the local Macedonian dialect as Pajakaški Vlasi after Mt. Pajak, where most Meglenoromanian villages are located.) Here we use South Danubian Balkan Romance (SDBR) as a cover term for Aromanian and Meglenoromanian when they can be treated together. The term Vlah is used only when quoting from another source.

The term Macedo-Romanian has also been used to refer to Aromanian or Aromanian together with Meglenoromanian in opposition to the term Daco-Romanian, which refers to Romanian (or Romanian and Moldovan, depending on time and politics). On the politics of Dacian identity in Romania, see Verdery 1991; cf. Dietler 1994 on France and the Gauls. In both cases a nation-state – or certain powers within it – that has a Romance official language has made strategic political use of the pre-Roman occupants. However, at present the term Macedo-Romanian implies that Aromanian is a dialect of Romanian (which, as a nation-state language, never takes a prefix, e.g., in published grammars of the language) rather than a separate language. On the other hand, it is worth noting that some Aromanian speakers in Romania refer to their language as macedonean ‘Macedonian.’ Istro-Romanian, spoken in Istria, left Romania sometime between the thirteenth and sixteenth centuries (Sărbi & Frățilă 1998: 35–43, but see also Filipi 2002) and is also sometimes given separate status, but it is not normally treated in accounts of Balkan linguistics. In a sense, the status of Istro-Romanian is like that of Arbëresh within Albanic (Friedman 2001b).Footnote ⁶⁶

Also in our discussion of East Balkan Romance we must address the question of Moldavian/Moldovan. The standard language of the former province of Bessarabia (the territory of the former principality of Moldavia between the rivers Prut and Dniester, which was ceded by Turkey to Russia in 1812, declared independence in 1917, joined Romania in 1918, became a Soviet Socialist Republic in 1944–1947, and declared independence as the Republic of Moldova in 1991), known as Moldavian (subsequently Moldovan) is based not on Bessarabian dialects but on the same Wallachian dialects as Standard Romanian (see Dyer 1996). There has also been vacillation in Moldova over whether to call the language Moldovan or Romanian. Thus, aside from factors resulting from Russian or Ukrainian influence, Literary Moldavian was in fact not an elaborated separate dialect but rather a form of Literary Romanian written in the Cyrillic alphabet until 1989, when Latin became the official alphabet. In 1994, the official language of Moldova was Moldovan.Footnote ⁶⁷ Since December 2013 it has been Romanian.

In terms of dialectal divisions, the main distinctions for Romanian proper are Moldavian (Moldova and Bucovina), Muntenian (Wallachia, southern Transylvania, and southern Dobrudja), Banat (including northeastern Oltenia), Crișana (including western Transylvania), and Maramureş (Caragiu-Marioţeanu et al. 1977). We can also mention here Beas (Boyash, Banjaš, etc.), a Romanian dialect spoken by groups of Romani descent (see also footnote 77) in Hungary, former Yugoslavia, Bulgaria and isolated groups in Greece and, perhaps, Albania, the archaisms of which indicate a separation from western (Banat, Crișana) dialects of Romanian in the eighteenth century (Sikimić 2005a, 2005b; on Albania see Weigand 1895: 78, see also Kahl 2012; Kahl & Nechiti 2012).Footnote ⁶⁸ In the case of Aromanian, although there are many subdivisions, according to Saramandu (1984: 427), a basic distinction can be drawn between the dialects of the north – especially of the Fărşerot, and Grabovean, of southern Albania and adjacent parts of North Macedonia and Epirus as well as Gopeştean and Maloviştean of North Macedonia, which are characterized by absence of a distinction between schwa and high-back-unrounded vowel (<ă> vs. <î> or <â> in Romanian orthography) – and the dialects of the south, especially Grămostean and Pindean of Epirus, eastern Macedonia, and Thessaly, which distinguish schwa from the high back unrounded vowel (cf. Kahl 2005; Saramandu & Nevaci 2006, 2014a). Owing to patterns of migration over the past two centuries, followed by the hardening of borders in the twentieth century, which altered or eliminated the traditional patterns of transhumancy, these differences are realized in North Macedonia as a west/east opposition. The southern group as represented by the Grămostenii, are found east of the Vardar, while west of the Vardar is a mixture of various northern groups. Thus, for example, Saramandu 1984: 427 distinguishes Bela (near Struga) as well as Gopeš and Molovište (Bitola region) as distinct within the northern group, but Bela is actually divided into two groups: the older Măbalot (from beala, via mbeala, ultimately from Moscopole, Alb Voskopoja) and the more recently arrived Fărshălot (see also Friedman 1994b).

The poorly attested, extinct Eastern Romance language, Dalmatian, like Istro-Romanian, does not generally figure in accounts of Balkan linguistics. Dalmatian is known mainly from word lists from the last speaker, who was from the island of Krk (Itl Veglia, hence also the name Vegliote for Dalmatian) and died in an accident in 1898 (Fisher 1976; Maiden 2004 and references therein). On the other hand, the Venetian dialect of Italian exerted an influence on the western and southern coastal Balkan peninsula comparable to that of Turkish in the interior.

Finally, there is Judezmo, which is the language brought by Hispanic-speaking Jews of the Iberian peninsula to the Ottoman Empire (and elsewhere) when they were expelled from Spain on August 2, 1492 and Portugal in 1496–1497.Footnote ⁶⁹ Coming to the Balkans at the invitation of Ottoman sultan Bayazid II, speakers of Balkan Judezmo spread out into mostly urban communities all around the eastern Mediterranean, in particular in Greece (especially the north and particularly in Thessaloniki and Kastoria, but also other towns (e.g., Véroia, Lárisa) and islands both in the Aegean Sea (e.g., Chios and Rhodes) and the Ionian Sea (e.g., Corfu), North Macedonia (especially Bitola, Skopje, and Štip but also Ohrid and elsewhere), Serbia, Bosnia-Hercegovina, Dalmatia, Bulgaria, and Turkey (most notably Istanbul and Izmir), and later (in the nineteenth century, coming mainly from Bosnia and Bulgaria) in Romania, too (particularly Bucharest). The language can be divided into two major dialects, Eastern and Western. Eastern Judezmo includes the dialects of Istanbul, Izmir, Rhodes, and Thessaloniki, while Western Judezmo includes the dialects of Belgrade, Sarajevo, Bitola, Bucharest, and Sofia. Most speakers of Judezmo were murdered in the Holocaust. Small remnant communities survive in Balkan towns (especially Istanbul and Thessaloniki but also elsewhere), Israel, and the United States. See Sala 1976 for a valuable and comprehensive bibliographic essay on the language. See also Quintana Rodríguez 2006, whose atlas provides a more complex and nuanced picture of various isoglosses. According to her, the northeast is more distinct from the west and southeast (Quintana Rodríguez 2006: 358).

1.2.3.4 Balkan Slavic

It is generally agreed that the Indo-European dialect that became Slavic acquired its recognizably Slavic shape in the northern part of Eastern Europe some time prior to the migration of Slavic speakers into the Balkans south of the Danube during the sixth and seventh centuries CE. On the basis of some old shared developments, South Slavic can be divided into two groups: East South Slavic and West South Slavic. The northernmost of the West South Slavic dialects became Slovene in what is today Slovenia and adjacent bits of Italy, Austria, and Hungary. There is linguistic evidence indicating that what became Slovene and what became Slovak (the southernmost of the West Slavic languages) remained in contact until the early tenth century CE, when Germanic and Hungarian speakers came between them approximately on the territory of today’s Austria and Hungary. The more southerly and larger group that became speakers of West South Slavic probably consisted of a single people, most likely the Slaveni, who at some point were divided and ruled by two military aristocracies, probably of Iranian origin, who have been identified as the sources of the modern ethnonyms Serb and Croat (see Fine 1983: 53–57 for details). Their dialects became the West South Slavic that eventually occupied the territory of today’s Croatia, Bosnia-Hercegovina, Montenegro, and Serbia (with later migrations to what became Italy, Austria, Hungary, Romania, and elsewhere). East South Slavic designates the Slavic dialects currently spoken on the territory of Bulgaria, North Macedonia, and adjacent parts of Greece, Albania, and southwesternmost Kosovo, but which we know to have been spoken all the way down to the tips of the Peloponnesian peninsulas, where a Slavic-speaking tribe known as Melingi were attested at least as late as the fifteenth century (Vasmer 1941: 18–19; Fine 1987: 166, 234). The evidence of toponymy also indicates the former presence of Slavic-speaking populations in other parts of what are today Albania and Greece (Vasmer 1941; Ylli 1997–2000).Footnote ⁷⁰

In connection with the conversion of the South and West Slavs to Christianity during the second half of the ninth century, the first known documents in a Slavic language were produced.Footnote ⁷¹ Byzantine missionaries from Thessaloniki, Methodius and his brother Constantine (who took the monastic name Cyril shortly before dying), are credited with having written or supervised the writing of these earliest documents, mostly translations of scriptures and other religious texts from the Greek, but also some translations from Latin and Old High German, and even some original compositions (Lunt 2001: 10). The originals have all been lost, and all that survive are later copies of a fraction of these texts, none of them older than the late tenth century, and most of them later. Based on the language of these copies, we can determine that in the late ninth century the various Slavic dialects, while already differentiated enough to be identifiable with certain regions, were nonetheless still mutually intelligible and not very far removed from what we can reconstruct as Late Common Slavic. The language of these earliest documents is called Old Church Slavonic, and it is defined as the non-East Slavic language of manuscripts and monuments that display certain archaic characteristics and are presumed or known to have been written prior to 1100 CE (see Lunt 2001: 1–14).Footnote ⁷² Documents dating from after this period and/or not preserving diagnostic archaisms are known as Church Slavonic (in various recensions) or as the “Old” stage of the various modern South Slavic languages.Footnote ⁷³

From the end of the Middle Ages to the beginning of the nineteenth century, the ancestors of what are now the modern South Slavic languages developed under different conditions in different regions and at various times. During these centuries, wars and population movements considerably complicated the dialectological picture of South Slavic. In terms of documentation, traditions varied from the vibrant vernacular literature of the Renaissance Dalmatian coast to the highly conservative but vernacular-influenced Church Slavonic ecclesiastical texts (damaskini) of Ottoman lands, the brief flowering of written colloquial language during the Reformation in Carinthia, Carniola, and Styria (of the fifty or so books published between 1550 and 1598, the Bible translation was especially significant), the mixture of Church Slavonic and Russian used by Orthodox Slavs in Hungarian lands, the vernacular literature composed by Slavic-speaking Muslims using Arabic script, and legal documents and chancery records of varied provenance, to name only some of the many types of written sources. In terms of the modern standard languages of today, however, in the Balkans, as in much of the rest of Europe, it was the rise of romantic nationalism and associated events and movements from the late eighteenth century onward that resulted in the current linguistic situation. Prior to that period, i.e., throughout the centuries during which the processes that led to the Balkan sprachbund were taking place, the names of speech forms and even the very concept of language were not isomorphic with our present concepts. (For an interesting discussion of the Western European developments that ultimately contributed to the construction of language that functions today in the Balkans and elsewhere see Bauman & Briggs 2003: 19–70.)

From a dialectological point of view, South Slavic linguistic territory presents an extraordinarily varied picture whose complexity can be captured only partially by the designation of isoglosses (cf. Alexander 2000a). Nonetheless, for heuristic purposes one can identify phonological and morphological developments whose geographic extent can be represented cartographically. The picture that emerges is a series of isoglosses that tend to cluster in certain areas but nevertheless do not define any radically sharp breaks (see Ivić 1958: 31, 32 and Friedman 1999a: 7). It is thus never the case that speakers from neighboring villages cannot understand one another, but as the differences increase with distance, eventually speakers from sufficiently separated villages will speak mutually unintelligible dialects, and those distances are less in some areas and greater in others. One group of isoglosses clusters in the region between what are today Slovenia and Croatia and another group clusters between the regions that today are located in southeastern Serbia and western Bulgaria and fans out across what is now the Republic of North Macedonia.Footnote ⁷⁴ During the course of the nineteenth century, a variety of projects for South Slavic standard languages were pursued with varying degrees of success. The details of these processes need not concern us here.Footnote ⁷⁵ As of this writing there are seven South Slavic official languages (moving roughly from northwest to southeast): Slovene, Croatian, Bosnian, Montenegrin, Serbian, Macedonian, and Bulgarian. There have been a number of attempts to develop standards from various other dialectal bases in all of these languages over the years, but for simplicity’s sake, we limit ourselves mostly to languages with titular nation-state support (but see below for some specific exceptions). Our task here is to determine how these relate to our object of study, the Balkan sprachbund.

Since the Balkan sprachbund, like all such contact phenomena, is the product of orality, the linguistic level with which we are concerned is the dialect. At the same time, however, space precludes the exhaustive exposition of every dialectal datum. To the extent that they are based on speech, standard languages provide a convenient tool for summarizing the relevant phenomena. South Slavic, however, is the only language group with contiguous dialects some of which are clearly Balkan and others of which clearly are not. The question therefore arises as to where to draw the line. In §1.1, we indicated competing definitions and claims relating to the borderland Balkan Slavic dialects as well as some of the terminological problems. In this book we refer to the Slavic dialects of Bulgaria and of Greek and Turkish Thrace as Bulgarian, the Slavic dialects of the Republic of North Macedonia, Greek Macedonia, and adjacent parts of Albania as Macedonian, and the Slavic dialects of the Republic of Serbia as Serbian or BCMS. The following exceptions are hereby noted: The dialects of the Bosilegrad and Dimitrovgrad (Caribrod) districts of the Republic of Serbia are acknowledged by both Serbia and Bulgaria as Bulgarian and the speakers consider themselves to be ethnic Bulgarians. The dialects of the Slavic-speaking Muslims of Gora, a region on the eastern and northern slopes of Mts. Korab and Šar in northeastern Albania and the adjacent southwesternmost corner of Kosovo are included with Macedonian unless otherwise specified.Footnote ⁷⁶ Macedonian-speaking Muslims are sometimes known as Torbeš (Alb Torbesh), a term that is also applied to Goran in North Macedonia and Albania. Some Muslim Macedonian speakers consider this term derogatory, while others embrace it. Serbian dialects in the Republic of North Macedonia are of two types. The dialect of the Gallipoli Serbs of Pehčevo near Berovo represents a Serbian enclave settled in North Macedonia from Turkish Thrace in the wake of the Balkan Wars, World War One, and the subsequent Greco-Turkish War (Ivić 1957: 5–6). Certain dialects in the north of the Republic of North Macedonia such as the village of Kučevište and other villages on Kozjak and Skopska Crna Gora north of Skopje are considered by their speakers to be Serbian because they identify as Serbian (Orthodox) and learn standard Serbian in school. These dialects, however, do not differ from the Macedonian dialects of the neighboring villages in terms of any structural features (Vidoeski 1998: 10). The dialects of Bulgarian-speaking Muslims can be specified by the term Pomak. At present Pomak is treated as a separate minority language in Greece (Kokkas 2004; Karakhotza 2006; Papadēmētriou 2013; Theocharidis 1996a, 1996b).Footnote ⁷⁷ For Bulgarian dialectologists, these Pomak dialects are part of the same Rhodopian dialect complex of Bulgarian spoken on both sides of the mountains/border (cf. Steinke & Voss 2007; Friedman 2012c). While there are some ethnolectal differences between the dialects of Christians and Muslims in the region (cf. R. Greenberg 1996a), the Slavic speakers of the Pirin Macedonia region (Blagoevgrad district of Bulgaria) are divided in terms of self-identification: some are Bulgarian-identified and some are Macedonian-identified. The Slavic dialects of Pirin Macedonia are therefore specified here as such. There are also Bulgarian enclaves in Romania and Moldova as well as Slavic-speaking Muslims in Turkey. We can also note here the presence of Serbian and Montenegrin dialects in northern Albania and a Bosnian enclave near Durrës and another enclave near Fier (Steinke & Ylli 2013).

1.2.3.5 Balkan Indic (Romani)

Like other modern Indic languages, Romani developed as a distinct language or dialect in India during the Middle Indic period, defined as 600 BCE–1000 CE (Masica 1991: 51), probably in Central India (see Matras 2002: 14–18, for a summary of the evidence). This means that in terms of attestations of earlier stages we have at our disposal the same data as for the rest of Indic: Vedic, Classical Sanskrit, Prakrits, and Apabhraṃśa. There is, however, a significant gap between these ancient languages and our first Romani documents, the oldest three of which are short lists of words and phrases dating from the sixteenth century (1542, pre-1570, 1597), all from England and Holland (Friedman & Dankoff 1991: 1; Matras 2002: 10). The fourth oldest Romani document, however, is from the Balkans, namely Komotini (Trk Gümülcine), in what is now Greek Thrace. It is a list of words and phrases with Ottoman Turkish translations in Evliya Çelebi’s travelogue, the Seyāhat-nāme (Friedman & Dankoff 1991). Based on linguistic evidence, it seems likely that Romani was in contact with Byzantine Greek in Anatolia by 1000 CE (Tzitzilis 2007b), although we cannot say for certain when the Roms crossed from Asia Minor to the Balkans other than to note that they were already in Constantinople by the middle of the eleventh century (Soulis 1961: 144).Footnote ⁷⁸ By the fourteenth century they had already begun their migrations to other parts of Europe.

Although Romani attracted the interest of some of the earliest Balkanists (Miklosich 1874–1878; Weigand 1895: 78), it was mentioned only to be excluded during the “classical” (or modern) period of Balkan linguistics. Thus, for example, Sandfeld 1930: 3 mentions Romani in a footnote without any attempt to integrate it into his findings, and Asenova 2002: 220 describes Romani as edin nebalkanski ezik, no s dălgo žitelstvo na Balkanite ‘a non-Balkan language but with longtime residence in the Balkans.’ Some individual studies of Balkan features and the Balkanization of Romani were published during the late twentieth century (Kostov 1973; Friedman 1985a; Matras 1994a), and Joseph 1983a does spend some, albeit limited, time on Romani in his monograph on the Balkan infinitive, but Romani has only recently begun to be integrated into any more extensive study of the Balkan languages. Thus, for example, Sawicka 1997 takes Romani into consideration, and Hinrichs 1999a contains two chapters treating Romani in its Balkan context (by Boretzky & Igla, and by Bochmann). This is in part owing to the general marginalization of Roms and in part to the lack of both adequate data and national structures for Romani. Among the reasons Romani provides important contrast to the rest of the Balkan languages is the unidirectionality of ordinary multilingualism, i.e., Roms learn other Balkan languages, but most non-Roms do not learn Romani. The same could be said about Judezmo.Footnote ⁷⁹

At present, the dialectological classification of Romani accepted by most linguists working on the language distinguishes four main groups: Balkan, Vlax, Central, and Northern (cf. Matras 2002: 5–13, 214–237). Each of these is subdivided into two groups apiece: Balkan I and II, Northern and Southern Vlax, North and South Central, and Northeastern/Northwestern.Footnote ⁸⁰ The two types of Balkan dialects are spoken by groups that have stayed, for the most part, in the so-called Southern Balkans (Albania, Kosovo, North Macedonia, Bulgaria, Greece, and Turkey) but also by groups that have migrated to Romania, Crimea, Azerbaijan, and Iran (Matras 2002: 6). The Balkan I dialects are more conservative and are spread throughout the area, while Balkan II are found in northern Bulgaria, Kosovo, and adjacent parts of North Macedonia and Albania. Of the former, the Arli (from Turkish yerli ‘local’) dialect is particularly widespread while of the latter, Bugurdži (from Turkish bürgücü ‘gimlet-maker’ also called kovači ‘blacksmiths’ in North Macedonia, as well as other names) is the predominant variety spoken in North Macedonia and Kosovo (Boretzky 1993, 2000b; Matras 2002: 223). The Vlax group, so named because of significant Romanian lexical influence presumed to have been acquired during an extended sojourn on Romanian-speaking territory, when certain shared phonological innovations developed, is divided into a Southern branch, which migrated back into the southern Balkans (Gurbet ‘migrant,’ Džambaz ‘horse-dealer,’ etc.), and a northern branch (Kalderaš ‘kettle-maker,’ Lovari ‘horse-dealer,’ etc.), some of whom remained in Romania, others of whom migrated south to northern Bulgaria, west to central Serbia, north to Hungary, Poland, and beyond, and many of whom joined the Eastern European emigration to North America in the late nineteenth and early twentieth centuries. The Central group is concentrated in former Austria-Hungary, the Northeastern in Russia, the Baltic lands, and central Poland, and the Northwestern in Germany, France, the Nordic countries, and, into the twentieth century, Great Britain (see especially Matras 2010).

The Balkan versus non-Balkan varieties of Romani are of particular interest also because Romani as such acquired its basic shape during its sojourn in the Balkans, and subsequent migrations have resulted in a differentiation between those dialects spoken in the Balkans, which continued to develop Balkanisms in contact with the other Balkan languages (e.g., future formation), and those dialects spoken outside the Balkans, which lack some basic Balkan features.

1.2.3.6 Balkan Turkic

The oldest Turkic monuments, the Orkhon inscriptions of the upper Yenisei, date from the eighth century CE and show remarkable affinities with Oghuz Turkic (Tekin 1968). As noted above (§1.2.1.11), we are interested here not only in Turkish as an adstratum, which has been the traditional approach, but also in Turkish as a participant in Balkan linguistic processes. That said, we can observe that although various Turkic-speaking peoples passed through or settled in the Balkans as noted in §1.2.1.11, we are concerned in this section only with the arrival of the speakers of the dialects that became Balkan Turkish in the Balkans. These dialects can be traced to the arrival of Oghuz-Turkic speakers in Anatolia in the eleventh century, a convenient date being the defeat of the Byzantine Emperor Romanus IV Diogenes by the Seljuk Turks under Alp Arslan at the Battle of Manzikert (Trk Malazgirt) in 1071, which opened Anatolia to Turkic conquest. Although – under pressure from the Kumans (Polovtsy) – a group of Ghuzz (Oghuz) Turks invaded the eastern Balkans from the north in 1064, most of them were wiped out by the plague and the rest scattered or became Byzantine mercenaries (Fine 1983: 211). As Fine 1987: 165 observes, by 1261, when the Byzantines recaptured Constantinople from the crusading Latins (“Franks”), who had held it since 1204, “Byzantium was hardly an empire any longer – despite its titles, rhetoric, and court ceremonial; it was just another petty state, holding, together with Constantinople, western Anatolia, Thrace, Thessaloniki, and Macedonia.” This set the stage for Ottoman expansion during the following century. By 1300, most of Anatolia was in Turkish hands, with Osman I, who established the Ottoman dynasty, ruling an emirate in northwestern Anatolia 1290–1326. During the mid-fourteenth century, various Turkish troops were used in Europe by rival Byzantine dynasties, and in 1352, a Turkish army defeated a Serbian one – each fighting for opposite sides in a Byzantine civil war – at Demotika (Fine 1987: 325–326). Technically, this was the first major Turkish–European battle in Europe. During this period, however, Turks in Europe were raiders and mercenaries rather than settlers. In 1354, taking advantage of an earthquake that had collapsed the walls of Gallipoli, the Turks crossed the Dardanelles and occupied the fortress for themselves, an event which marked the beginning of their occupation of Europe as a political force. Adrianople (modern Edirne) fell in 1369 (Fine 1987: 406). The decisive defeat of Serbian forces occurred at Chernomen (Grk Orménio) on the River Marica in 1371 (Fine 1987: 379) – near what is today the Turkish–Greek–Bulgarian border – although the later Serbian defeat at Kosovo Polje in 1389 is more famous. By the end of the fourteenth century Ottoman rule covered all of what would become Balkan Slavic linguistic territory as well as what is today eastern Greece and southern Romania, and by the end of the fifteenth century Ottoman rule had expanded to include the entire region with which we are concerned here.Footnote ⁸¹ The core of the Balkan linguistic area as we have defined it remained under Turkish rule until the early twentieth century, although the peripheries began to assert their independence in the nineteenth. Even after the retreat of the Turks to eastern Thrace, however, Turkish remained a language of urban sophistication and prestige, especially in those countries that did not expel their Muslim populations (i.e., it remained so everywhere except in Greece).Footnote ⁸² Turkish was still spoken by town dwellers in North Macedonia regardless of religion well into the second half of the twentieth century (VAF, field notes,1973–2001) and it remains an important language for Muslims throughout the Balkans (cf. Ellis 2003 on North Macedonia).Footnote ⁸³

The Turkish dialects of the Balkans are divided into two groups, East Rumelian and West Rumelian (Németh 1956). The location of the boundary between the two is remarkably similar to that of the so-called jat-line (see Stojkov 1968: 56), a major isogloss distinguishing East Bulgarian from West Bulgarian (Hazai 1961). This is to say that the local Turkish dialects of Kosovo, Macedonia (including Greek Macedonia before 1923; cf. Mollova 1960), and Albania, as well as western Bulgaria (Kakuk 1960), are all of the West Rumelian type, except for the dialect of the Yürüks of eastern North Macedonia, who are more recent arrivals and speak an East type of Rumelian Turkish (Nedkov 1986; Jašar-Nasteva 1986; Manević 1953/1954). East Rumelian is basically an extension of Istanbul Turkish, while West Rumelian shows considerably more Balkanization at all levels of its grammar (Ibrahimi 1982; Friedman 2002a; cf. Katona 1969; Kowalski 1926). Gagauz has been classed as an East Rumelian dialect with Indo-Europeanized (i.e., Balkanized) syntax (Menz 1999), and in this respect is also of interest here.

1.2.3.7 Language Choice and Dialect versus Standard

Owing to the fact that the different members of a given linguistic group frequently display the same Balkanisms/Balkan linguistic phenomena (ceteris paribus and mutatis mutandis), it is generally the practice to cite one or at most two representatives from any given group, usually in standard orthography, with dialectal examples supplied only where particularly relevant. Sandfeld generally uses dialectal examples, in part because he was relying on folklore texts or on works published before the relevant languages had been standardized, and in part because some of the languages about which he was writing had not yet been standardized at the time he was writing.Footnote ⁸⁴ All of the handbooks of the late twentieth and early twenty-first century have followed the practice of citing mostly standard forms taking Bulgarian as representative of Balkan Slavic and Romanian as representative of Balkan Romance. While such a practice is not in and of itself misrepresentative, as long as appropriate data from other languages and various dialects are cited where relevant, in this book we nonetheless follow a different practice based on the centrality of Macedonia for Balkan sprachbund phenomena in general. (For a discussion of Macedonia as the “heart” of the Balkan sprachbund, see Hamp 1977a; Topolińska 2010.) Except where the specificity of the data requires it, we take Standard Macedonian as our representative of Balkan Slavic, Standard Albanian as our representative of Albanic, and Aromanian as our representative of Balkan Romance. Since Standard Macedonian is based on its west central dialects and Standard Albanian is based on northern Tosk, the dialectal bases of the respective standard languages are at the heart of the heartland. Moreover, Aromanian represents the form of Balkan Romance with which the other languages have been in contact for the longest time. Since the standardization of Aromanian is still in progress (see Friedman 2001b), we take Gołąb’s 1984a grammar of the Kruševo (Aro Crushuva) dialect and Markovikj’s 2007 monograph on the dialects of the Ohrid-Struga region as basic (these dialects being among those that have been in contact with the dialectal bases of both Standard Macedonian and Standard Albanian). For Greek, we use the Demotic standard but with reference to northern or other dialects as appropriate. For Romani, we take the Arli dialects of Skopje as our base, these being both a particularly representative dialect and the base of standard Romani as used in the Republic of North Macedonia (the only country in the world to mention the Romani people (Romi, romskiot narod) in its constitution).Footnote ⁸⁵ For Turkish, we use standard or West Rumelian as appropriate.