6.0 Introduction: The Question of Morphosyntactic Balkanisms
As we have already observed (§1.1, §3.5), the term Balkanism can be used for any linguistic development that can be plausibly linked to the historical multilingual contact that characterizes the Balkan sprachbund. Nonetheless, morphosyntactic Balkanisms have constituted the heart of modern Balkan linguistic study. In the context of our approach to Balkan linguistics, we understand morphosyntax to be the realm that involves inflectional morphology, the passage of inflectional morphology into analytic paradigms and syntactic constructions, and the passage of syntactic constructions into inflectional and analytic paradigmatic ones. These are the phenomena that constitute the subject of this chapter.Footnote 1 Chapter 7 then takes up the deployment of these phenomena within discourse structure, i.e., syntax and, in certain respects, pragmatics. Owing to the fluidity of such heuristic boundaries in real human languages (“All grammars leak,” in Sapir’s 1921: 38 famous formulation), Chapters 6 and 7 taken together cover the phenomena treated solely under morphosyntax in many other works, while Chapter 7 also includes phenomena that are strictly syntactic in nature, e.g., involving the ordering of elements.
It was precisely certain Balkan morphosyntactic features that first attracted the attention of linguists in the nineteenth century, although elucidating the origins of Balkan vocabulary occupied more scholarly energy until the end of World War Two.Footnote 2 Thus, for example, Miklosich’s 1862 pioneering list of morphosyntactic Balkanisms (§2.2.3) took up two pages of a seventy-page article on the Slavic lexicon in Romanian, and the largest chapter (pp. 16–99) of Sandfeld’s 1930 classic synthesis was devoted to vocabulary. By contrast, Asenova 2002 devotes 216 pages to morphosyntax and only thirty-three pages to lexicon. Nonetheless, while it is true that Sandfeld 1930: 100–162 devotes more space to bilateral concordances entre différentes langues balkaniques en dehors de lexique (‘convergences among various Balkan languages outside the lexicon’) than to concordances générales en dehors de lexique (‘general convergences outside the lexicon’) (Sandfeld 1930: 163–216), those two chapters taken together surpass that on the lexicon and actually contain observations, often the earliest, of most of today’s known morphosyntactic Balkanisms (broadly understood). Much of the work since Sandfeld has refined, expanded, and discovered new sources for phenomena that he noted in one form or another. Thus, for example, double determination, i.e., the use of a demonstrative with a definite form, is noted by Sandfeld 1930: 122 in his section on Albanian-Slavic correspondences (although Sandfeld also observes that Greek has the same type of construction), but we can add that in fact the feature is pan-Balkan insofar as it also occurs in Balkan Romance and Romani, albeit not necessarily pan-dialectal for all the Balkan languages. Similarly, the future-in-the-past as conditional, described in greatest detail by Gołąb 1964a and Belyavski-Frank 2003, is already noted by Sandfeld 1930: 105 in his section on Greek-Aromanian correspondences, where he also mentions Macedonian, Bulgarian, and Tosk Albanian.
As can be seen in the following sections (especially §§6.1.1, 6.1.1.4.8, 6.1.5, 6.2.1.3, 6.2.4.3.2, inter alia), the oft-cited tendency toward “Balkan analytism” is an oversimplification and to some extent an overstatement too. While the replacement of nominal case and nonfinite verbal inflection moves in that direction, the creation of analytic paradigms from syntactic constructions complicates morphology in ways that have frequently been overlooked. The classic example of this is the emphasis so often placed on the loss of case in Balkan Slavic while at the same time the elimination of inherited verbal categories in North Slavic, especially Russian, is ignored (Friedman 2011e). The Russian verbal system in particular is considerably impoverished with respect both to that inherited from Common Slavic and that which developed in Balkan Slavic, which not only preserved all the inherited distinctions but, in most dialects, also elaborated on them and developed new ones, as is seen below (§6.2.4.2 and §6.2.4.2.1, and cf. Hacking 1997a). Aronson 2007 has even gone so far as to argue that Russian is grammatically aberrant among the Slavic languages in its degree of restructuring – with passing reference to the Uralic substrate hypothesis (cf. Veenker 1967; Dahl & Koptjevskaja-Tamm 2001; Dombrowski 2013) – and that Balkan Slavic is more truly “European” (but cf. §3.4.1.3, where we argue against the Eurological approach to the history of the Balkan sprachbund).Footnote 3 We can also note here the fact that Balkan Romance is more conservative than non-Balkan Romance in its preservation of case inflection (cf. §6.1.1).
As in most traditional approaches to morphosyntactic Balkanisms, the best recent example of which (aside from the shortcomings noted in §3.5, Footnote footnote 156) is Asenova 2002, this chapter looks at the nominal system and then the verbal system. There is a methodological justification for this approach insofar as the restructuring of Balkan nominal systems involves more of the kind of morphosyntactic innovation that was first noticed as diagnostic of the sprachbund by Leake 1814 and Kopitar 1829 as well as in Trubetzkoy’s 1923, 1930 original formulation. Our discussion of these two systems is organized by grammatical categories, rather than by Balkanisms per se.
6.1 Nominal Morphology and Morphosyntax
All the Balkan languages have simplified their inherited patterns of nominal inflection in terms of case marking, which affects the marking of nominal relations such as possession, partitivity, location, direction, and argument role. Although some simplifications began prior to the arrival of the various relevant languages in the Balkans, (and such simplifications have occurred in Indo-European languages outside the Balkans as well), a number of specific developments took place during the period of Balkan language contact and with specifically Balkan results. While case inflection tended toward reduction (§6.1.1), the morphological or morphosyntactic expression of reference underwent elaboration, particularly in marking for definiteness and indefiniteness, with deixis also having a role to play, albeit more limited (§6.1.2). In fact, the elaboration of referentiality is connected to the maintenance of case in many instances, and if pragmatic considerations such as focus and topicality are taken into consideration as well, then the continued role of case marking is even more apparent. In gender distinctions, such general simplifications as occurred in the Balkan Indo-European languages (the loss or retreat of the neuter) were probably already in place or well under way either before contact occurred (Romani, Albanian) or were already underway before the break-up of the language group (Romance). Nonetheless, some syncretisms (and, perhaps, conservatisms) relating to grammatical gender or noun class can be identified as having Balkan relevance (§6.1.3).
Grammatical number shows very little in the way of Balkan contact-induced change aside from the infiltration of some plural markers and the spread of the T/V distinction (Brown & Gilman 1960; Friedrich 1972), which latter is probably more Eurological than specifically Balkan (§6.1.4).Footnote 4 If we include numerals and numerical expressions under the general rubric of number, then the phenomena in question are basically lexical (§4.2.2), although the collocation used for teens (and twenties in Aromanian) does involve the lexicalization of ancient morphosyntax. Insofar as Hamp 1992a is correct in postulating a pre-Balkan, Northwest Indo-European origin for the ‘on-ten’ construction of teens, however, then the areal origin of the resemblances is prior to Balkanization.Footnote 5
For adjectives (§6.1.5), the change from synthetic to analytic gradation is definitely of Balkan relevance, and the infiltration and adaptation of nonagreeing adjectives can also be mentioned. The question of adjectival order, like that of head-genitive relations, is more properly the realm of syntax (§7.3.3), although insofar as adjective order interacts with definiteness marking, morphosyntax is arguably involved.
We consider contact-induced pronominal change (other than that covered in §6.1.1.2.2 on possession and §6.1.3.3 on agreement) in the section on lexical borrowings with grammatical value (§4.2.3.1). In this kind of division of labor between lexicon and grammar, we are following research such as King 2000 (used by Labov 2007). While contact-induced grammatical change in the Balkans makes it clear that the power of the lexicon as a conduit for grammatical change should not be exaggerated (as was done in Labov 2007: 349), Balkan pronominal systems have either been relatively conservative, or they have undergone changes that can be treated in one of the abovementioned sections.Footnote 6
6.1.1 Case, an Overview: Loss, Maintenance, and Analytism
The starting point for cases in all the Indo-European Balkan languages was an eight-case system consisting of nominative, accusative, genitive, dative, locative, instrumental, ablative, and vocative.Footnote 7 The tendency to simplify, merge, and lose substantival case categories long pre-dates the arrival of the earliest forms of each of the attested relevant languages in the Balkans, although some had gone further than others by the time they arrived. In this section, we give a synopsis first of these case-related developments, and then offer more details on a language-by-language basis.
Owing to the fact that the vocative has the potential to refer outside the narrated event to participants in the speech event, i.e., to involve an addressee, it is sometimes treated as not being a case in the sense of the relational cases, which are all bound to the narrated event. For our purposes, however, the fact that the vocative is a part of nominal inflection is crucial in assessing its declensional status. While the vocative often does refer to a participant in the speech event, it is not limited to that function, insofar as it can reference the participant in a narrated speech event. We do not have more to say about the vocative in this section, but return to it in §6.1.1.4, as it occupies an important place in Balkan case developments. On the one hand it serves as a counterpoint to some tendencies, and on the other hand it is consistent with some general trends.
In this section we also omit pronouns – where case categories are generally more robust – despite some signs of loss and some movement in the same direction as nominal trends. The generally more conservative nature of pronouns vis-à-vis case is seen in other languages in the family, such as English and French – to be taken as representative of trends in Germanic and Western Romance more generally – with nominative-accusative distinctions in personal pronouns (I/me, je/me, respectively) but not in nouns.
The basic facts concerning cases in the Balkan languages as they are usually presented in the handbooks focus on standard languages, but they constitute a useful point of departure for our discussion. Albanian shows a well-developed case system, with nominative, accusative, genitive, dative, and ablative cases (although the genitive and dative are identical except for the particle of concord in the former; the locative is now dialectal), although there is syncretism in many categories, and some distinctions are made only in definite forms. Standard Bulgarian has no overt case distinctions in the substantive, signaling some case relations analytically with prepositions, especially na for both genitive and dative.Footnote 8 The same is basically true for Macedonian, although remnants of oblique marking on masculine proper names and a handful of animate nouns – a western dialect feature – is part of the norm. Animacy marking on nouns in Macedonian, however, has not been taken up by younger speakers of the standard and is, in modern practice, a regionalism. Balkan Romance shows a case system with nominative, accusative, genitive, and dative, but various mergers have reduced this to essentially a nominative-accusative case and a genitive-dative case. Greek has a system with nominative, accusative, and genitive, with genitive fulfilling some dative-like functions, and syncretisms affecting the realization of cases, such that most masculine nouns merge accusative and genitive while feminine nouns merge nominative and accusative. Finally, Romani shows a full panoply of cases, with all of the Proto-Indo-European case distinctions – nominative, accusative, genitive, dative, instrumental, locative, and ablative, and even the vocative – realized, even if the forms are not directly inherited but rather reflect renewals based on new markers.Footnote 9 Turkish has nominative, accusative, genitive, dative, locative, and ablative, which represents a reduction of the Old Turkic system (Erdal 2004: 167–182), and Judezmo had already participated in the Hispanic loss of case prior to arrival in the Balkans. Thus Balkan case systems show a mix of case loss, case syncretism (merger), analytism in the place of case, and retention with renewal.
Going now into more detail, especially of a historical nature, we turn first to Albanian. Owing to the absence of relevant attestation, we do not know the chronology of case loss for the language that became Albanian. Sh. Demiraj 1985: 249–297/2002: 189–217 gives the most thorough survey of various theories. Except for Bokshi 2005: 317–474, who argues that Albanian completely lost the IE case system and then restructured a new one, various scholars are agreed that Albanian preserves remnants of PIE cases. The processes of simplification, however, are quite old. Given the distinct ablative plural and remnants of a locative in a variety of dialects extending from northern Tosk to the southern part of Northern Geg and including some of Arbëresh (Gjinari 2007: Map 193), we can argue that the order of loss was instrumental first, then locative, the current locative being an innovation. The current Albanian indefinite ablative plural may in fact be a genitive survival, although this question remains moot. Nominative and accusative are identical for indefinite forms, but distinct in the definite singular, and while genitive and dative are merged in form in the singular and plural, they are distinct in their syntax, the dative co-occurring with a reduplicated object pronoun (see §7.5.1) and the genitive with the particle of concord. These mergers are evident in the oldest Albanian documents.
For Greek, complete replacement of ablative by genitive had already occurred before the Mycenaean period. The locative and instrumental were completely replaced by the dative sometime after the collapse of Mycenaean civilization (c.1200 BCE) and before the return of literacy to Greek (late ninth century BCE; Horrocks 2010: 9–10). In Classical times, the fact that the dative had replaced the instrumental and locative led to ambiguities and the emergence of alternative expressions such as διά + gen for instrumental (a very early example of the tendency for the genitive to replace the dative – see §6.1.1.2 below). The replacement of the dative by the accusative in locative expressions (ἐν + dat > εἰς + acc) was well advanced by the first century CE, and by the tenth century CE, the accusative was the only case occurring after prepositions in the spoken language. Examples in late medieval texts of prepositions plus genitive must be taken to be intentional archaisms (Browning 1983: 36–37, 82–83). The dative was replaced by accusative in the north (cf. Thumb 1912: §54) but also in other dialects, and by the genitive in the south, a process which was completed by the fifteenth century CE but undoubtedly began considerably earlier. The accusative for indirect object is attested already in the Greek of the so-called Proto-Bulgar inscriptions of the ninth century CE (Horrocks 2010: 284; see §6.1.1.1.3.1 for details; note that these inscriptions are named for their writers, not for their language). On the other hand, as discussed in greater depth in §6.1.1.1.3.2, use of εἰς ‘to’ + acc for indirect object NPs is already attested in the Greek New Testament.
For Romance, the instrumental had disappeared completely and the locative was already obsolescent in the Old Latin period (up to 75 BCE).Footnote 10 Prepositions gradually took over the functions of other case endings as shown by various Late Latin inscriptions (Mihăescu 1978: 43–50). Thus, for example we see cum ‘with’ plus accusative as well as ablative, ob ‘for’ with accusative, ablative, or genitive, and ad with accusative was replacing the dative around the beginning of the Common Era, i.e., before Latin arrived in the Balkans. Nominatives were replacing obliques, de was marking the partitive genitive, and adnominal datives were replacing genitives, all by the second century CE (Mihăescu 1978: 322). Evidence from Latin inscriptions in the Balkans ends in the sixth century CE, but by then we already have evidence of nominative for accusative and ablative and vice versa in Dalmatia (Mihăescu 1978: 217). By the time Balkan Romance appears after the millennium-long hiatus, the case system is more or less as we see it today. The only cases remaining are nominative, accusative and dative-genitive (from the older dative), except for Meglenoromanian, which has a genitive distinguished from a dative by the use of different connecting particles (e.g., in some villages lu with genitive versus la with dative; see Atanasov 2002: 213–214).
For Slavic, the ablative and genitive had merged in the Balto-Slavic period. Unlike in Greek, however, it was the ablative, not the genitive, that was generalized.Footnote 11 Aside from this one merger, Common Slavic, as attested, mutatis mutandis, in Old Church Slavonic, arrived in the Balkans with the remaining PIE case distinctions intact, although there are hints of simplification already in the oldest documents (Wahlström 2015).Footnote 12 Such conservatism has become emblematic of a putative “Slavic linguistic type” in modern times, no doubt influenced by Russian nationalism in its pan-Slavic guise. Still, Balkan Slavic is the only Indo-European language group that is attested in the Balkans at such a late date with most of the original PIE case distinctions. While dative possessives are already attested in the oldest documents, as is accusative for locative after a preposition (e.g., in Zographensis, Koneski 1981: 161), case merger did not begin in earnest (at least as attested in writing) until the late medieval period (twelfth to thirteenth centuries), and the process was not complete in what we can call the heartland until the eighteenth to nineteenth centuries.Footnote 13
“Heartland” here refers to the zone of Balkan Slavic where case-loss was most complete in contrast to all those regions peripheral to it that preserve remnants of substantival declension. In the west of Balkan Slavic territory, the heartland zone consists of Macedonian dialects east and south of an isogloss following the Bregalnica and Crna rivers, continuing southeastward into Greece (Vidoeski 1998: 93–95).Footnote 14 The zone extends into Bulgarian dialect territory, excluding the northwesternmost dialects transitional to Serbian, and continues eastward approximately to the river Vit in the northeast and the river Văča in the southeast. East of that region, declensional remnants again occur all the way to the Black and Aegean Seas, except for most of Strandža (including the adjacent corners of Greek and Turkish Thrace) and Istanbul province.Footnote 15 In Turkey and Greece, the zone of preservation extends from Tekirdağ province across Greek Thrace and into Aegean Macedonia as far west as Sérres.Footnote 16
In all the peripheries of the heartland of total substantival case loss, the most common retention is oblique case marking (based on the old animate accusative) for male proper names and kinship terms and, sometimes, other animate nouns, e.g., Stojan/Stojana. North of an isogloss bundle running north of Tetovo and south of Kumanovo and Kriva Palanka, feminine accusatives also appear, e.g., sestra/sestru ‘sister’ (nom/acc), and west of a bundle that starts at the border with Kosovo and runs southwestward, west of Kumanovo, cutting through the west central Macedonian dialect region – east of Brod and Resen and west of Prilep and Bitola – dative masculine and feminine names and kinship terms occur, e.g., tetin/tetinu ‘uncle (aunt’s husband)’, tetka/tetke~tetki ‘aunt.nom/dat’ (Vidoeski 1998: 65–95).
At the extreme, mountainous peripheries in Gora (Mladenović 2001: 269–275), we have full paradigms of N(ominative)-A(ccusative)-D(ative)-V(ocative) for both animate and inanimate substantives as well as adjectives, although the accusative is distinctive only for masculine animates. The dative, however, is consistently distinguished.Footnote 17 As one moves north and east into the Torlak BCMS dialects, we find the three-case (N-A-D) plus vocative completely preserved.Footnote 18 If one continues northward, declension becomes more robust until we have the full system of standard BCMS. If, however, one continues eastward, there is gradual erosion back to personal masculine only as one moves toward the Bulgarian heartland zone.
Many of the dialects in the Rhodopes and Thrace have a full N-A-D declension, at least for masculine animate singular nouns, as seen in the Pomak paradigm from Greek Thrace (Adamou 2010a) (see Table 6.1).Footnote 19
| ‘father’ | nom.sg | bubajko | nom.pl | bubajkove |
| acc.sg | bubajka | acc.pl | bubajkove | |
| dat.sg | bubajku | dat.pl | bubajkovem |
Such remnants of case marking must be viewed against the backdrop of the emergence of definite marking, as in some dialects, most cases are marked on the article or are triggered by it (see also §6.1.2.2.1.1). The paradigms from the dialect of Boboshtica (Mac Boboščica), in the Korça region of Albania, based on Mazon 1936: 54–64, are illustrative (see Table 6.2).
| ‘the old man’ | nom.sg | starec-o | nom.pl | starci-ti |
| acc.sg | starca-togo (indf starec) | acc.pl | starci-ti | |
| dat.sg | starcu-tomu (indf starec) | dat.pl | starci-tim | |
| ‘the sister’ | nom.sg | sestra-ta | nom.pl | sestrjä-te |
| acc.sg | sestra-ta | acc.pl | sestrjä-te | |
| dat.sg | sestrjä-tuj | dat.pl | sestrjä-tem | |
| ‘the village’ | nom.sg | selo-to | nom.pl | sela-ta |
| acc.sg | selo-to | acc.pl | sela-ta | |
| dat.sg | selu-tomu (indf selo) | dat.pl | sela-tam |
As observed by Mindak (1987/1988), the parallels to Albanian in Korča Macedonian are an archaism probably encouraged by contact with Albanian. Note also, however, from Mazon, nom Petar, acc Petra, dat Petru ‘Peter,’ nom/acc brat, dat bratu ‘brother,’ but ot brata-mi ‘from my brother,’ nom/acc.sg kniga, dat.sg/nom.acc.pl knige ‘book.’ Case marking can also be conveyed on adjectives and demonstratives, e.g., parvetemu starec ‘the.first old.man.dat,’ sijam starci ‘these old.men.dat.’
Šklifov 1973: 63–64 notes datives of the type car ‘king,’ carutomu ‘the.king.dat.’ Even some neuter domestic animals can take the dative definite article, although they do not themselves decline, e.g., magare ‘donkey,’ magaretomu (dat). Interestingly enough, while the dative has been recorded in Kostur (Grk Kastoria), the animate accusative is absent. According to Šklifov, these usages are found mostly in folktales and phraseological expressions.Footnote 20
With regard to other cases in Balkan Slavic, the locative was the first to be lost, a merger with the accusative that also occurs in Montenegro (Koneski 1981: 162). Then the accusative replaced the instrumental (a process also attested in the twelfth to thirteenth centuries).Footnote 21 In the final stages, dative completely replaces genitive (a process already attested in the oldest documents), with both ultimately coming to be marked prepositionally with na. But the animate accusative based on the genitive – itself a late Common Slavic innovation that was not fully developed at the time Slavic arrived in the Balkans – survives on the periphery and, moreover, replaced the inanimate accusative in the pronoun in all of Slavic. We can also note here the generalization of 3pl gi from accusative to dative in Sretečka Župa (Pavlović 1939: 173).
Thus, for Balkan Slavic especially, it must be emphasized that while case-loss is widespread within Bulgarian and Macedonian, at the same time the peripheries (except the Lower Vardar dialects of Macedonian) all show instances of case-retention. At the same time, however, the movement towards analytism is continuing in western North Macedonia, affecting pronouns, an otherwise conservative area of the grammar, as well as masculine personal nouns. Thus, for example, nonclitic dative pronominal objects are spreading from the east westwards within Macedonian, marked by na rather than an inflected form, e.g., na mene, na nas, na nego, na kogo rather than western (and officially preferred standard) mene ‘me.dat,’ nam ‘us.dat,’ nemu ‘him.dat,’ komu ‘whom.dat,’ etc.Footnote 22 There is even a further innovation with the question word, na koj in the meaning of ‘to whom?,’ with the nominative form of the pronoun, consistent with a move towards the reduction or even elimination of case marking.Footnote 23 This tendency to eliminate case inflection continues to move westward, affecting not only Macedonian usage but also Aromanian (see Markovikj 2007: 49–94). Also the animate oblique, which is facultative in the Macedonian norm, is now rare and has become a western regionalism. In Bulgarian, according to Stojanov 1983: 413, the animate oblique is perceived as an archaism. All other case relations are likewise indicated syntactically throughout Balkan Slavic, usually by a preposition but sometimes just by juxtaposition (see §6.1.1.3).
Romani shared in the Middle Indic loss of Old Indic declension and the rise of postpositional constructions while still on the Indian subcontinent.Footnote 24 By the time Romani arrived in the Balkans, however, (and perhaps already in Byzantine Anatolia), it had become a prepositional language, and the surviving postpositions (plus a bit of older material) had petrified into an eight-case system recapitulating the distinctions attested in Sanskrit (Friedman 1991). The difference between Romani and, e.g., Slavic, in this regard is that Romani recapitulates but does not preserve the Proto-Indo-European system of case marking per se whereas in Slavic, at the time it arrived in the Balkans, the case inflections were direct descendants of the Indo-European ones. Romani’s case system has remained fairly conservative throughout its speech area, although in the late twentieth and early twenty-first centuries the case system has begun to retreat and be replaced with prepositional constructions. Despite a lack of empirical studies demonstrating a correlation between increased Romani integration, education, and literacy, on the one hand (which, despite the many problems, nonetheless rose dramatically after 1989), and certain types of innovation, on the other, it can be hypothesized that Romani’s case system is a site of change in its Balkan context at least in part owing to shifts in social relations (see §6.1.1.5.1).
As far as Turkic is concerned, Erdal 2004: 182 identifies 11 “active and productive” case suffixes for the Old Turkic substantive: nominative, accusative (replaced by the pronominal accusative), genitive, dative, locative, ablative, instrumental, equative, directive, partitive-locative, and similative. A twelfth case, the comitative, is neither active nor productive. Pronouns have fewer case distinctions than nouns. Orkhon Turkic had nine cases (Tekin 1968: 125). By the time Turkish arrived in the Balkans, however, the current system of nominative, (definite) accusative, genitive, dative, locative, and ablative was already in place, the other case relations being handled by postpositions or suffixes that, while of declensional origin, are no longer active and productive (in Erdal’s sense).
The changes that affected substantives also affected other parts of speech. Of greatest importance are the pronouns (sensu lato, thus including demonstratives that feed into definiteness marking), which, in their functions as reference and person markers become the primary bearers of case relations and are of considerable importance to the discussion of case. Adjectives, insofar as they become bearers of referential marking, participate in these pronominal developments, but otherwise adjectives tend to show fewer remnants of case than substantives.
In the sections that follow, accordingly, the relevant developments involving nominal morphosyntax are presented, organized around cases and their fates. The vocative is included here (§6.1.1.4) based on the argument advanced above concerning its importance for understanding the full Balkan picture, and attention is given as well to the absence of any overt case marking (zero marking, §6.1.1.3), as it too is relevant to the overall view.
6.1.1.1 Accusative Developments
In discussions of the accusative case, analytics such as subject/object, agent/patient, actant roles, etc., when viewed through the prism of Balkan linguistic contact-induced convergence, can be summarized in terms of three Indo-European basics: direct object, indirect object, and other nonsubject.Footnote 25 In the Balkans, the Indo-European accusative (sensu lato) has merged with the nominative in the substantive in Balkan Romance (everywhere), Albanian (in the indefinite), and most of Balkan Slavic (except, depending on dialect, virile, ±masculine, ±feminine sg), Greek (except m), Romani (except animate).Footnote 26 In the pronoun, the accusative maintains its identity, and in some instances goes on to absorb the dative.
6.1.1.1.1 Accusative as the Casus Generalis after Prepositions
A specifically Balkan feature with the accusative is the generalization of the accusative after prepositions, giving a so-called casus generalis ‘general case’ (see also §6.1.1.3 on zero marking). Its Balkan nature is shown by the fact that it is characteristic precisely of Balkan Slavic vis-à-vis the other Slavic languages and even other languages and dialects within South Slavic. Although the process is attested sporadically in the earliest documents, it must be remembered that even the oldest of these are not originals, and that the originals (as well, perhaps, as some of the oldest copies) date from the ninth century CE, i.e., at least two centuries after the arrival of the Slavs in the Balkans. As noted above, the simplification took almost a millennium to complete in the central zone and is still incomplete at the peripheries. Similar developments can be traced in Balkan Romance and in Greek, but Albanian and Romani are quite different.
As for Balkan Romance, the Latin starting point was a system in which prepositions were generally differentiated as to which of two cases, accusative or ablative, they could govern, e.g., ad ‘to,’ inter ‘between,’ per ‘through,’ trans ‘across,’ etc. with accusative and ab ‘away from,’ dē ‘down from,’ ex ‘out of,’ cum ‘with,’ etc. with ablative, and some prepositions could take both cases, with an associated difference of meaning, e.g., in + accusative ‘into’ versus in + ablative ‘in.’Footnote 27 The present-day situation is well illustrated by Romanian, where one still finds multiple case possibilities with prepositions, but of a quite different kind. That is, all of the basic, i.e., underived, prepositions govern accusative case, whereas “the prepositions that select the genitive or the dative are always derived (from nouns, adverbs, or participles)” (Mardale 2013: 534). The former type includes all those inherited from Latin, including a ‘like,’Footnote 28 cu ‘with,’ de ‘of,’ în ‘in,’ pe ‘on,’ and others, whereas the latter type includes recent (and still obvious) derivatives, e.g., graţie ‘thanks to’ (+ dative), from the noun graţie ‘grace,’ or mulţumită ‘thanks to’ (+ dative) from the participle mulţumită (cf. a mulţumi ‘to thank’), and transparent composites such as în faţa spitalului ‘in front of the hospital’ (cf. faţă ‘face’).
Greek has a similar pattern to Slavic and Romance in the Balkans in showing a move towards accusative as the casus generalis with prepositions, except that given its long recorded history, it is possible to see how this development unfolded over millennia. That is, the seeds of the move towards accusative-only prepositions can be seen in the sporadic reduction of case-government possibilities in Ancient Greek. Bortone 2010 gives a detailed look at the full history of developments with prepositions, and gives examples (pp. 156–157) of genitives in place of the usual dative case with παρά meaning ‘near,’ e.g., παρ’ ὑγρῶν Ἰσμηνοῦ ῥείθρων ‘by the wet currents of the Ismenus’ (Soph. Ant. 1123) and of accusatives with the same παρά, e.g., ἦν παρὰ τὴν ὁδὸν κρήνη ‘there was a fountain by the side of the street’ (Xen. An. 1.2.13); the former can be seen as an early encroachment of the genitive on the dative and the latter as presaging, as it were, the rise of the accusative as a casus generalis with prepositions. He points out further (pp. 203–205) that by Hellenistic Greek, it was rare to find cases other than accusative with prepositions, and ultimately (p. 434), in Medieval Greek the accusative is established as the case that occurs in prepositional phrases.
In fact, from Medieval times on into Modern Greek, there are really only four basic (primary) prepositions (Joseph & Philippaki-Warburton 1987: 201): locatival-directional σε (earlier εἰς) ‘in, into, to, at, on,’ benefactive-purposive για ‘for,’ comitative-instrumental με ‘with, by,’ and ablatival από ‘from, by,’ though some (e.g., Thumb 1910: 105) would add κατά ‘towards, according to,’ αντί(ς) ‘instead of,’ ως ‘to, up to, till,’ and χωρίς/δίχως ‘without.’ Along with these, there are numerous so-called “improper” prepositions (“compound” or “complex” for Joseph & Philippaki-Warburton 1987: 204), which, as Thumb 1910: 107 describes them, “denote mostly spatial relations” and consist of adverbs “converted into prepositions by a genitive coming after them,” or in combination with one of the basic prepositions (see also Qvonje 1991: 30ff.). However, the only genitives possible are the weak genitive pronouns, and only the first four basic prepositions listed here, not Thumb’s additional ones, can combine. Thus, these combinations have the peculiarity of directly governing as their objects genitives only of weak pronouns, but otherwise accusatives headed by a basic preposition. Examples include μπροστά μου / μπροστά σ’ εμένα / μπροστά στην πόρτα ‘in.front of.me’ / ‘in.front of (to) me’ / ‘before (to).the door,’ κοντά του / κοντά στην Αθήνα ‘near him’ / ‘near to.Athens,’ μαζί μας / μαζί με τους φίλους μας ‘(together-)with us’ / ‘(together) with the friends of.us,’ πίσω σου / πίσω από το σπίτι ‘behind you’ / ‘behind (from) the house,’ and αντί για τον Γιάννη ‘instead of (lit., ‘for’) (the) John.’ Parallel instances can be cited from Medieval Greek, giving an indication of the age of the combinations; Karatsareas 2016: 57 gives the examples in (6.1):
a.
μέσα ἀπὸ τοὺς ἵππους within from the.acc.pl horse.acc.pl ‘(They took his body) (out) from among the horses’ (Troy (thirteenth/fourteenth century), 5070) b.
μαζί με τ’ ἄρματα σας together with the.acc.pl chariot.acc.pl your ‘(I want your horses) together with your chariots’ (Achilleid O (sixteenth century), 163)
Moreover, Karatsareas gives tables (pp. 54, 56) for Medieval Greek with, respectively, essentially the basic prepositions of Thumb (minus ἀντίς) and the adverbials Thumb mentions that combine to form improper prepositions. Thus this system with accusative as the casus generalis with prepositions was firmly in place by early Medieval Greek.
As for the other languages, they stand outside of the zone formed by Slavic, Romance, and Greek within the Balkans. Albanian is definitively not Balkan in this respect, since prepositions can govern the nominative (nga ‘from’ and te[k] ‘at’ (in the sense of Frn chez)) itself a specifically Albanian development.Footnote 29 Remaining prepositions take the accusative or ablative (which in most, but not all, instances is identical with the dative/genitive).Footnote 30
In Romani, most prepositions take the nominative for substantives and the dative or locative for pronouns, but the privative bi ‘without’ usually takes the genitive (Boretzky & Igla 2004: Map 33), and animates can take the accusative, e.g., ko kakes ‘to the uncle.’Footnote 31 There is some dative-locative merger (often in favor of the dative). Bi- also functions as a prefix and thus has special status. We should also note that the replacement of case forms by prepositional constructions is, unsurprisingly, characteristic of substantives more than pronouns. (Cf. also Turkish postpositions with nominative substantives and genitive pronouns, e.g., Ahmet için ‘for Ahmet.nom’ benim için ‘for me.gen.’) It is, however, arguably a contact phenomenon rather than an internal development. Friedman 1991 discusses the development of Middle Indic prepositions into what is now a case system (see also Matras 2002: 92–94; Elšík & Matras 2006: 218–238). In the Balkans, some dialects have borrowed the Turkish post-nominal positioning of adpositions and show optional postpositions: Rhodopian Romani katar o voš alongside o voš katar ‘from the forest’ (Igla 1997: 153, cited in Matras 2002: 92); see further §4.3.3.2 on the borrowing of adpositions. In some dialects, Turkish postpositions are borrowed as prepositions but with the Turkish case government, e.g., Turkish sonra ‘after’ borrowed into what is now Agía Varvára Romani in a colloquial form, sona, governing the ablative: sona duj čhonendar ‘after two month.pl.abl’; cf. Turkish iki aydan sonra, lit., ‘two month.abl after’ (Messing 1988: s.v.). Gilliat-Smith 1915/1916: 87 (cited in Friedman 1991) records the use of the Turkish postposition beri ‘until, since’ as a postposition used both with and without the copied Turkish ablative case marker dan, e.g., siknara-beri or siknara-dan-beri ‘since childhood.’ Similarly, Haskovo Romani has a native form aškal ‘according to’ used postpositionally with the dative, e.g., e dasenge aškal ‘according to the Bulgarians’ (lit., ‘the Bulgarians.gen according.to,’ Ivanov 2000). Although the dative, locative, and ablative are frequently replaced by calqued prepositions in North Macedonia, Matras 2002: 93 reports that in Greece, borrowed me ‘with’ replaces instrumental in Romani dialects in Greece (see also RMS for various types of examples).
6.1.1.1.2 Preposition for DO
In both Balkan Romance and in some Macedonian dialects, the morphosyntax of marking a direct object is or can be achieved by the use of an ostensible preposition.Footnote 32 Consider, for instance, the following examples from Romanian, the language in which the construction is best instantiated (adapted from Pană Dindelegan 2013: 129ff.):Footnote 33
a.
L- am rugat pe tata him.acc.wk have.1sg asked dom father.def.acc ‘I have asked father’ b.
Te ajut pe tine you.acc.wk help.1sg dom you.acc ‘I help you’ c.
Îl caută pe profesor him.acc.wk search.3sg dom teacher.acc ‘He is looking for the teacher’
In general, pe, which otherwise means ‘on,’ is found in this construction occurring with no lexical meaning but marking as objects personal animate nouns that moreover are specific; note the contrast in meaning between (6.2c) and (6.3):
(6.3)
Caută menajeră search.3sg housekeeper.acc ‘He is looking for a housekeeper’ (i.e., ‘any housekeeper’ not a specific one)
This construction with pe, according to Pană Dindelegan 2013: 131ff., “places Romanian among the languages with a strongly marked direct object. The fact that the prepositional construction is related to other features (‘high’ placement on the animate and on the specificity scale), places Romanian … in the category of languages that display ‘Differential Object Marking’” (‘DOM,’ in the sense of Bossong 1985).
This prepositional construction is quite normal in Romanian, and is discussed in all available grammars of the language. As for Aromanian, while it occurs in some Aromanian-speaking communities, e.g., Ohrid-Struga (Markovikj 2007: 72, 87), Turia (Grk Krania; Bara et al. 2005: 43) as well as other dialects in both North Macedonia and Greece (Asenova & Aleksova 2008: 14–15), it is not systematically registered everywhere where it might be expected. In Meglenoromanian, as best can be determined, it does not occur: Hill 2013 expressly says as much, and Atanasov 2002: 267 gives sentences in Romanian, the meta-language for the book, with prepositional marking of the object as translations of Meglenoromanian sentences with no such marking; for instance, he translates ou vizui Cana ‘I saw Cana’ into Romanian as am vazut-o pe Cana, with the DOM marker pe.
As for Macedonian, where it occurs, the construction uses the preposition na, which corresponds to pe in the meaning ‘on.’ It is documented in mostly peripheral areas concentrated in western Macedonia. Koneski et al. 1968: 521 note the occurrence of “la construction avec la préposition na pour l’objet direct” (‘the construction with the preposition na for the direct object’) in the southwestern Macedonian region, and specifically in Debar, Kostur (Grk Kastoria), and Kajlar (Grk Ptolemaḯda), and Koneski 1986 finds it in the nineteenth century in the Kulakia (Grk Khalástra) gospel and notes it in Struga as well. Speakers also consider it characteristic of Gostivar (VAF field notes). Sobolev 2008 and Topolińska 1995a locate it similarly, as do Asenova and Aleksova 2008. Adamou 2006, 2010a reports it for the Macedonian of Ajvatovo (Grk Liti), near Solun (Grk Thessaloniki).Footnote 34 Examples are in (6.4), adapted from the indicated sources:
a.
Go vidov na Petreto (Koneski et al. 1968) him.wk.acc saw.1sg dom Petre.def ‘I saw Petre’ b.
Jas gu vidu na deteto (Bužarovska 2001) I him.acc saw.1sg dom child.def ‘I saw the child’ c.
jisk-at da a-sfalj-at na moma-ta want-3pl dms her.acc-descend-3pl dom girl-def ‘They want to make the girl come down’ (Adamou 2010a)
The situation with Aromanian is complex and needs some extra attention here. Koneski et al. 1968: 521 state simply that it is “également courante en aroumain” (‘equally current in Aromanian’) and give as an example ‘l vidzui pi Taki ‘him I.saw dom Taki,’ where pi is one of the Aromanian equivalents of Romanian pe, along with pri, pre, and prə. Similar examples are to be found in Markovikj 2007: 70 et passim for Bela di Suprã Frasheriote Aromanian in North Macedonia, as in (6.5a), and in Bara et al. 2005: 45 for Turia (Grk Krania) Aromanian in the Pindus area in Greece, as in (6.5b):
a.
U vãdzuj pi Toma kum ini dupã mini. him.wκ.acc saw.1sg dom Toma as come.3sg after me ‘I saw Toma as he was coming after me’ b.
nu-lu kunusk’u prə elin’opulu, prə fič’orlu neg-him.wk.acc know.1sg dom Greek.child dom boy ‘I do not know him, the Greek child, the boy’
And, some instances are to be found in Papahagi 1974, under the entries for pi (p. 971) and pri (p. 1010), as Sobolev 2008: 117 points out.
On the other hand, Pană Dindelegan 2013: 138 refers categorically to “the absence of the pe structure” in Aromanian, as do others, as noted by Sobolev 2008: 117–118, e.g., Caragiu-Marioțeanu 1975: 237, and Atanasov 2002: 80. Moreover, Vrabie 2000: 67, in discussing prepositions in Aromanian, states overtly that Romanian uses pe “to introduce direct objects in the accusative” but says nothing about Aromanian other than observing that the Aromanian sentences he gives with Romanian translations are “without the preposition pre or pri (equivalent to [Rmn] pe).” And, Gołąb 1984a: 78–81 says nothing about the use of pi (or pre/pri) in his discussion of cases in Kruševo Aromanian. Asenova and Aleksova 2008 speculate that the influence of Romanian schools in Macedonia in the late nineteenth and early twentieth centuries might have been relevant.
Sobolev 2008: 116–117 decries this contradictory situation in the literature and tries to understand what it means by noting that “the lack of prepositional direct object is typical for only a part of Aromanian dialects, like that of Kruševo … other dialects of Aromanian, like that of Kranea/Turya … do really use the prepositional strategy to mark the direct object.” However, Asenova & Aleksova 2008: 1 also have examples from Kruševo. As a result, one has to be careful about blanket statements about the language, and even specific dialects, as precisely this key point of morphosyntax is subject to dialectal, or even idiolectal, variation. Thus, it is clear that some Aromanian speech communities – perhaps even many – have this structure. There is a question of the extent to which the conditioning is grammatical or pragmatic, but it can be taken as an established fact that DOM with pi, etc., exists in Aromanian, even if it is not found across all of Aromanian territory or in every context in a given dialect. And, where it does occur, the examples show that this construction in Aromanian is restricted to personal animate direct objects (including pronouns).
To round out the presentation of Romance in the Balkans, it is worth pointing out that Judezmo also shows the prepositional marking of direct objects, also restricted to personal and animate nouns. In the texts provided by Wagner 1930: 95, the examples in (6.6a) occur, and Symeonidis 2002: 197 offers (6.6b), among others (orthography has been modernized; see ):
a.
¿krees tú ke io kero a padre manko bien? … yo amo believe.2sg you that I love.1sg dom father enough well I love.1sg muncho a padre much dom father ‘Do you believe that I love father well enough? …I love father greatly.’ b.
no tenemoz yana de mirar a ningunos neg have.1pl appetite comp see.inf dom anyone ‘We don’t have the desire to see anyone.’
As these examples show, the preposition in the Judezmo case, as in the relevant Western Romance languages, is a, which otherwise means ‘to,’ not the Balkan Romance pe/pi/pri/prə.
This fact signals an important point about the history of this prepositional DOM construction. It has a deeper history on the Romance side that cannot be ignored. Within Romance, it is found, quoting Pană Dindelegan 2013: 130, in “Catalan, Spanish, Portuguese, Sardinian, some Franco-Provençal and Italian varieties,”Footnote 35 and in all of those the preposition is a, from Latin ad ‘to, towards.’ That distribution in itself might suggest a Vulgar Latin source, but it is telling that a different prepositional source is found in Eastern Romance, namely Latin per ‘by; through’ (> pe, pi, pri, etc.). While the prepositional construction could reflect a Vulgar Latin structure with one or the other preposition, it may have arisen independently in each language; as Pană Dindelegan 2013: 130 puts it, these languages “have gone through the same type of innovation, marking the direct object prepositionally … only the preposition differs.” Here it is worth noting that semantically, lexical pe corresponds to Macedonian na in its locational meaning.
In the Balkans, it is clear that Judezmo, in the use of a, shows its Hispanic roots and that it brought the a construction with it when it entered the Balkans. But would the independent-innovation view go so far as to suggest that the Balkan Romance instantiations were independently created in Romanian and Aromanian? Admittedly, it is hard to say: Is the dialectally restricted occurrence of the construction in Aromanian the result of the receding of a structure inherited from Proto-Balkan Romance, and does its absence altogether in Meglenoromanian show the end point of such a recession? Or did Meglenoromanian never have it at all, and the inner-Aromanian distribution shows an innovatively emerging construction? Or, as noted above, was this a Romanianism imported during the nineteenth century spread of schooling? We leave these as unanswered, and perhaps unanswerable, questions.
Still, there is some evidence concerning the diachrony of the prepositional construction within Romanian itself. Starting from the reasonable speculation that there may have been some discourse-functional motivation involving distinguishing among arguments with certain verbs, Pană Dindelegan 2013: 135 summarizes the early history as follows:Footnote 36
Most researchers agree upon a date of emergence later than the separation of the dialects … but prior to the 16th century, when the construction is attested in Romanian. The quasi-general absence of PE in translated religious texts of the 16th century, on the one hand, and its occurrence in original texts, on the other, has been explained through the fact that the structure appeared little before this date. The construction had not yet fully established itself; therefore PE could be suppressed in translated texts, as an effect of the absence of the PE-construction in the original text.
Hill 2013 further observes that while the use of pe appears to be optional in early Romanian texts, the Moldavian chronicles of the seventeenth to eighteenth centuries, it is actually “the means to achieve a Contrastive Topic reading on the direct object” (p. 145). She notes too that “the classes of nouns to which [pe-DOM] applies have shifted considerably. In the Moldavian Chronicles, pe-DOM occurs with both animates and inanimates, mass nouns or count nouns.” In the modern language, as noted earlier, pe mostly marks animate personal specific nouns. Hill argues that “this switch reflects the switch in the discourse function of pe-DOM, from a marker of Contrastive Topic, as needed, to a backgrounding device for the [+human] noun in the discourse” (p. 147). Moreover, she links this switch to a connection between pe-DOM and object reduplication (see §7.5.1), in that in early Romanian, object reduplication was a backgrounding process while pe-DOM was a foregrounding process. Nonetheless, they came to overlap, first with pronouns, “where their reading had to be kept ‘familiar,’ as opposed to emphatic/Contrastive” (p. 148). In fact, pe-DOM seems to have spread rapidly through the system of strong pronouns “from third to first and second persons” and soon became obligatory, going together with a change in categorial status of pe. The modern situation shows some obligatory uses of pe but some that are discourse-sensitive, reflecting perhaps the discourse origins of the construction and some degree of syntactic generalization.
On the Slavic side, contact with Aromanian has been assumed by several scholars, including Koneski 1986 and Sobolev 2008, which more or less squares with the shared animacy restrictions and with the dialect distribution, since Aromanian is (or was) found throughout most of the Macedonian territory where this construction is found. Moreover, it is consistent with the fact that this construction occurs exclusively in the western (and in some instances southwestern) Macedonian periphery and not at all elsewhere in Balkan Slavic. That is, precisely where Aromanian has or had its presence, this type of construction appears in the local Slavic dialects, and not elsewhere in Balkan Slavic. The possibility of such influence even in Ajvatovo (Grk Liti) and Kulakia (Grk Khalástra) cannot be ruled out, given the Aromanian presence in Thessaly.
On the other hand, Topolińska 1995a: 95 suggests that Greek is the source, despite the fact that Greek influence as far north as Debar or Gostivar is difficult or impossible to justify; however, she does not provide details as to how Greek might have influenced Macedonian in this regard. Bužarovska 2001 develops this suggestion into a viable account. She draws on the fact that northern Greek dialects, which would generally have been in contact with the relevant Macedonian speech communities (but also, it must be noted, the Aromanian), use the accusative as the case for direct object but also, innovatively from the Greek point of view, for indirect object, as in (6.7):
(6.7)
τον έδωσα το βιβλίο him.wk.acc give.pst.1sg the.acc book.acc ‘I gave him the book’ (cf. Standard (southern) Greek: του έδωσα …, with gen του)
The equivalence of marking in northern Greek for indirect and direct objects, according to this version of the argument, would have allowed Macedonian speakers to innovatively extend their indirect object marker na into use as a marker of direct object. Debar and Gostivar, however, remain problematic for this account.
Moreover, this account does not explain the animacy restriction but rather has to rely on the fact that indirect objects – though not direct objects – are generally animate; nor does it account for the dialect distribution, which mirrors the current or former presence of Aromanian speakers more than that of Greek speakers, as noted above. Thus for western Macedonian, Aromanian influence seems to be a more compelling explanation. Moreover, Koneski 1986: 202 demonstrates, from a consideration of nineteenth-century Macedonian material, that the na-construction is well entrenched in southernmost Macedonian dialects by the middle of that century. Bužarovska 2020 surveys the other historical evidence, and concludes that it made its way into dialects around Kostur (Grk Kastoria) by the end of the nineteenth and beginning of the twentieth centuries, and has been spreading northward; while these are areas that historically had concentrations of Aromanian speakers, some spread within Macedonian itself is also possible.
However, Adamou 2010a argues that Aromanian cannot be invoked as the source of the prepositional marking for Liti Macedonian. If we ignore the presence of Aromanian south of Liti and assume that there was never any Aromanian-Macedonian contact in the region despite this fact, then either Greek is responsible or else we have to reckon with independent innovation, driven by the sorts of functional motivations that lead speakers to innovate DOM constructions in numerous languages. The problems with attributing the innovation to Greek have been noted above; however, appealing to typological universals raises the specter of such an account for all instances, including all of Balkan Romance and all Macedonian dialects that show this construction. Perhaps a mix of internal and external forces, as suggested by Bužarovska 2020, is the right approach. Crucial here is precisely the areal distribution for Balkan Slavic, viz. the limitation to the southwestern periphery of Macedonian, a region where Aromanian influence is already significant and was significant as well in the past (cf. Gołąb 1984a135, 1997).
In the end, there may be no good answer here, but the facts themselves are interesting and are worthy of the Balkanist’s attention. That this construction gives the Balkans relevance to the typologically and theoretically rich arena of DOM could simply be considered as a bonus.Footnote 37
6.1.1.1.3 Accusative/Dative Developments
The accusative case in some of the Balkan languages has spread at the expense of the dative case or of other cases that have replaced the dative in various functions. That is, if we conceive of certain typical dative case functions, however they are expressed, we can see that accusative forms can be involved in supplanting them. It is necessary to couch the discussion of these particular instances in these terms owing to the merger of dative and genitive forms and functions that characterize the nominal case systems of various Balkan languages, as detailed in §6.1.1.2 below.
For personal pronouns, the Balkan languages almost always distinguish accusative from dative in the 3rd person, but not necessarily in the 1st and 2nd. Thus, for example, Albanian merges the two in both long and short 1/2sg/pl forms. In Bulgarian, only the 1sg/2sg weak object pronouns have distinct forms.Footnote 38 Moreover, in eastern Rhodopian Bulgarian, dative can replace accusative in 1pl/2pl pronouns (Iliev 2017: 9–10). In Macedonian, only the 1sg/2sg long forms display merger, although that situation is currently undergoing change (see §6.1.1.1.3.1). In Torlak BCMS, as represented by Vratarnica (Sobolev 1994: 177), the 1pl/2pl long forms fall together in favor of the accusative. In general BCMS does not exhibit syncretisms. In Romanian, there is partial syncretism in the 1pl/2pl weak forms (the dative has both syncretized and distinct forms), and in the 3pl gender is distinct in the accusative but not the dative (Vasilescu 2013: 381–382). For Aromanian, there is merger in all three persons in the short form plurals, and there is gender merger in the 3sg short accusative and dative the 3pl long and short forms (Markovikj 2007: 69). In Meglenoromanian, 1sg/2sg long and 1pl/2pl show gender but not case merger, while 1/2sg clitics have distinct accusative and dative forms. In the 3rd person, the long forms are syncretic, but the short forms are not (Atanasov 1984: 205–206). For Greek, 1sg/2sg weak object pronouns distinguish accusative and dative (usually referred to as genitive for historical reasons), as does the third person in general except the 3pl masculine form (Holton et al. 1997: 95–97). Neither Romani nor Turkish show such syncretisms in terms of the accusative and dative forms of pronouns, but Romani does use the accusative where a dative might be expected, e.g., in possessives and as the indirect object of del ‘give.’
In those instances where something happens to the dative case vis-à-vis the accusative (see §6.1.1.2 regarding the dative vis-à-vis the genitive), two types of development can be noted. On the one hand, there are mergers of accusatives and datives (or dative-successors) in at least some functions, and on the other, there are combinations of various prepositions with accusative nouns that serve functions previously expressed by datives (or successors to the dative case). In the former type, accusative is directly involved in the merger while in the latter type it is only indirectly involved, with a preposition playing the key role from a functional standpoint.
6.1.1.1.3.1 Merger to Accusative Case Alone
Aside from the pronominal systems noted above, the first type of development, with accusative case alone marking the indirect object, is really found only in Greek and, to some extent, Romani (see also §6.1.1.1.3.2), and in its most extensive form is dialectally restricted as noted already in §6.1.1. However, despite this restriction, the development is noteworthy as it is consistent with the overall Balkan scenario of case reduction that occurs within all of the languages to different degrees (see §6.1.1). Moreover, the location of the relevant dialect area suggests that contact may well have been involved in this Greek development.
It is in the northern dialects of Greek that the bare accusative case is found most extensively in the marking of indirect object function, occurring in place of the genitive – the modern successor to the ancient dative – that is found in other dialects, especially those in the south. Importantly, in the north, this syncretism of genitive and accusative, in favor of the accusative, is found just for indirect objects and not for the possessive function of the genitive. Moreover, this syncretism is found for pronouns as well as nouns. Thus, for instance, where the south has (6.8ab), with, respectively, the first person genitive pronoun μου, and the genitive singular noun του Πέτρου for the indirect object, the north has (6.9ab), with, respectively, the accusative με and accusative τον Πέτρο:Footnote 39
a.
μου δίνετε τα λεφτά? me.gen give.prs.2pl the.acc.pl money.acc.pl ‘Are you giving the money to me?’ b.
δίνετε του Πέτρου τα λεφτά? give.prs.2pl the.gen Peter.gen the.acc.pl money.acc.pl ‘Are you giving the money to Petros?
a.
με δίνετε τα λεφτά? me.acc give.prs.2pl the.acc.pl money.acc.pl ‘Are you giving the money to me?’ b.
δίνετε τον Πέτρο τα λεφτά? give.prs.2pl the.acc Peter.acc the.acc.pl money.acc.pl ‘Are you giving the money to Petros?’
This same type of accusative IDO construction is used in Romani with both pronouns and proper names, e.g., ka de man pandž dinara (Jusuf 1974) ‘[you] will give me 5 dinars’ O Kurteši del e Ramo gayreti (Jusuf 1974) ‘Kurteš gives encouragement to Ramo’ (lit., ‘def.m.nom K. give.3sg.prs def.m.acc R. encouragement’). Romani also has accusative for possession, as in si man duj čhave ‘I have two children/sons,’ i daj si la duj čhave ‘the mother has two children/sons’ (lit., ‘the mother is her.acc two children/sons’) where the accusative pronoun la is marking a relationship handled by dative-genitive in other languages (cf. §6.1.1.1.3.3). The possessive construction is inherited from Indic (Matras 2002: 174), and accusative for IDO is widespread in Romani, and there are accusative-dative mergers in South Asian Indic, albeit of different sorts (Masica 1991: 365–367). Thus, while interesting as partial parallels to northern Greek, they are arguably independent developments.
Returning now to Greek, given that the IDO use of the accusative is characteristic of northern dialects, and that what is now the northern regions of Greece were once largely populated by Slavic speakers, it is tempting to think in terms of contact with Slavic as somehow being responsible. Such a position is conceivable, though perhaps not all that compelling. The reasoning behind such an attribution would have to involve the following, drawing on both the early Slavic use of the genitive for animate masculine direct objects, as found in OCS and to this day in BCMS (and elsewhere in Slavic),Footnote 40 and the later possibility of expressing possessives and indirect objects in Balkan Slavic with a prepositional phrase headed by the preposition na (on which see §6.1.1.1.3.2), as outlined in Table 6.3:
| I. | Slavic acc = gen for some nouns, a subset of those most likely to be indirect objects (animates) | |
| + | II. | Slavic gen can be expressed as na (+ Noun) for possession |
| + | III. | Slavic na (+ Noun) also is used for Indirect Object |
| ∴ | IV. | (Theoretically) acc could be used for Indirect Object |
This chain of reasoning is plausible but far from impeccable as a suitable historical account. First, Slavic did not actually instantiate (IV) with accusative used for indirect object functions (except insofar as the acc became the casus generalis after prepositions). It is just a theoretical possibility that one could be led to given the facts of (I) through (III). Thus, this chain would have to be a way in which Greek speakers learning Slavic made sense of the Slavic system that they then imported into their native Greek (reverse interference – see §3.2.1.3) or a way in which Slavic speakers learning (shifting to) Greek approached the Greek system, even if they did not realize it as such in their own language. Second, the starting premise, (I), is a generalization about the form of a direct object whereas the conclusion, (IV), is an extension on the functional side of the abstract case category of accusative. The mixing of the formal and functional domains makes this less than compelling. A final and probably decisive problem is that there are potential issues with chronology, and specifically whether the na + Noun pattern co-existed with a stage where the case forms for Balkan Slavic were still viable; see §6.1.1.1.3.2 on the dating of the Slavic pattern. On the Greek side, while Horrocks 2010: 284 sees accusative for indirect object in Greek as “a clear north/south choice [i.e., division] by no later than the fifteenth century,” importantly, as noted in §6.1.1, he points also to examples as early as the beginning of the ninth century, in the Proto-Bulgar inscriptions written in Greek by the Bulgar (Turkic) conquerors of Slavs in what is now Bulgaria; for instance, (6.10) occurs in an inscription dating from 813 (Horrocks 2010: 326):
(6.10)
τα δε λυπὰ κ[ά]στρα ἔδωκεν ο θε[ό]ς the.acc.pl and remaining.acc.pl town.acc.pl gave.3sg the.nom.sg god.nom.sg φόβον fear.acc.sg ‘And God gave (to) the remaining towns fear’
It is hard to know just what to make of this. It does show that accusative-for-dative was earlier in Greek than the usual chronology of the north-south split based on the characteristic northern vowel developments; Horrocks 2010: 405, for instance, dates the northern features of High-vowel Deletion (original i/u > Ø) and Mid-vowel Raising (original e/o > i/u) to after 1100 (see §§5.4.1.5, 5.4.3.9). And, this usage occurs in a variety of Greek that would have had some contact with Slavic. However, Horrocks voices the important caveat that the usage in the Proto-Bulgar inscriptions in general might not represent that of native speakers. While the innovative usage of accusative for earlier dative could not have been a calque on Turkic usage, the power associated with a ruling military elite could certainly have been a factor in promoting the spread of its version of Greek.
More problematic, perhaps, is the fact that accusative indirect objects are found in dialects of Greek that had no contact with Slavic. In particular, this usage occurs in Cappadocian and Pontic Greek (Horrocks 2010: 284; Thumb 1912: 39); Thumb (p.3 07) gives the following Pontic example:
(6.11)
είπεν το λεοντάρ τον πάρδον (Text III.13) say.pst.3sg the.nom lion.nom the.acc cat.acc ‘The lion said to the cat’
Moreover, it is found even in the southern dialects in the strong personal pronominal forms; that is, taking Standard Modern Greek as representative of nonnorthern dialects,Footnote 41 we find accusative-genitive merger, in non-3rd person forms (see Table 6.4).
| Singular | Plural | ||
|---|---|---|---|
| 1st | acc | εμένα | εμάς |
| gen | εμένα | εμάς | |
| 2nd | acc | εσένα | εσάς |
| gen | εσένα | εσάς |
The same holds in the weak object pronouns in the plural: μας (1st, acc/gen), σας (2nd, acc/gen) but not the singular (e.g., 1sg.acc με vs. 1sg.gen μου). These facts suggest that accusative for indirect object may have been an early tendency in at least Postclassical Greek that came to be realized most robustly in the north. Horrocks 2010: 284 sees this as an entirely Greek-internal process, a reasonable position to be sure, inasmuch as there were verbs in Classical Greek, such as κελεύειν ‘to order, command’ that could take either an accusative or a dative of the person ordered, presumably a notional indirect object, suggesting the roots of this tendency may reach even farther back in the history of Greek. Still, it is possible that contact with Slavic with its theoretical possibility of accusative indirect objects, enhanced that tendency. Nonetheless, it seems that contact alone cannot be responsible for this use of the accusative in Greek.Footnote 42
6.1.1.1.3.2 Prepositional Realization of Dative
More widespread in the Balkans is the second type of development, in which the dative or dative successor is expressed by a prepositional phrase, generally one with an accusative nominal as object. This usage is most evident with regard to the marking of indirect objects, though it extends to other uses of the older dative case.
As far as indirect objects are concerned, this development is found relatively early in Greek, as noted in §6.1.1, occurring in the New Testament (Rcvd) a couple of times with the preposition εἰς ‘to, in, into’ governing an object in the accusative case, as in:
a.
[οὕτως γέγραπται] … κηρυχθῆναι … μετάνοιαν … εἰς πάντα τὰ ἔθνη thus write.3sg.pass proclaim.inf.pass repentance.acc to all.acc.pl the.acc.pl nations ‘[It is thus written] repentance (is) to be announced to all nations’ (Luke 24: 47) b.
ἐκηρύξαμεν εἰς ὑμᾶς τὸ εὐαγγέλιον τοῦ θεοῦ proclaim.aor.1pl to you.pl.acc the.acc.sg gospel.acc.sg the.gen.sg god.gen.sg ‘The gospel of the lord is proclaimed to you’ (1Thessalonians 2: 9)
However, it must be noted that this usage, involving “εἰς of the recipients of a message,” has been branded by Blass & Debrunner 1961: 112, §207 as a Semitism, based on Aramaic l’ ‘to.’ Still, even if a Semitism, it is a usage that fell on a willing audience, so to speak, since Classical Greek shows some instances of the use of prepositions, including possibly εἰς, that are very close in sense to indirect objects. In particular, (6.13) from Thucydides (with a variant ἐς of εἰς) is suggestive:
(6.13)
ἐσπέμπει γράμματα ἐς βασιλέα (Thuc. 1.137) send.in.prs.3sg letter.acc.pl to king.acc.sg ‘He sends a letter to the king’
In this example, ἐς (= εἰς) is a variant reading for πρός, which with an accusative object means ‘(directed) toward’; Smyth 1920: §1695c gives this sentence with ἐς and notes πρός as the variant reading; editions of the text, e.g., Jones & Powell 1942, in the Oxford Classical Text series, and Smith 1928, in the Loeb Classical Text series, give only πρός here. However, if Smyth’s endorsement means anything, then εἰς/ἐς was a possibility in Classical Greek as a marker of an ostensible indirect object. And even if πρός is the only form to be taken seriously here, it still demonstrates prepositional marking for a notional indirect object.Footnote 43While such prepositional constructions gain currency, the dative survives in Greek until the tenth century, according to Humbert 1930. In fact, Humbert 1930: 161ff. states that it is the grammatical use of the dative, i.e., as a marking for indirect objects, that survived the longest of all of the various uses of this case. Thus, while there are language-internal processes at work with the rise of prepositional indirect object marking in Greek, the overall process played out over a very long period of time, a period during which contact with speakers of other languages increased.
Similarly, in Latin, while the dative case was the usual case of the indirect object,Footnote 44 an alternative prepositional structure with “ad ‘to, into,’” governing an accusative, can be found as early as Plautus; Salvi 2011: 339 cites (6.14), and suggests that with ad, prototypically “an indicator of place[,] … the sense of place could prevail and so there was ad with persons too”:
(6.14)
hunc … ad carnificem dabo (Plautus Captivi V.4.22) this.acc.sg to executioner.acc.sg give.fut.1sg ‘I’ll give this fellow to (into the hands of) the executioner.’
The prepositional construction with ad was clearly a part of Vulgar Latin since it is found all across Romance; cf. Spanish a, French à, Italian a, etc. Salvi 2011: 342 states that both the prepositional construction and the (bare dative) case construction “coexisted over a long period, even after case distinctions … fused into a single oblique case form” (in non-Balkan Romance languages).
In Balkan Romance, both strategies are to be found, inasmuch as the dative case has been retained (see §§6.1.1, 6.1.1.2), but actually the case-marking possibility is somewhat restricted. It occurs only in Romanian, as in (6.15), from Mihail 2013: 150:
a.
El a terminat de povestit copiilor. he has finished DESUP tell.sup child.def.dat.pl ‘He finished telling stories to the children.’ b.
Ion dă mâncare animalelor. Ion give.3sg food animal.def.dat.pl ‘Ion feeds the animals.’ (lit., ‘Ion gives food to-the-animals’)
It is absent from Aromanian, where, as Vrabie 2000: 48 puts it, the genitive/dative case form “is always preceded by the invariable particle a,” which continues Latin ad. However, alu and ali occur with singular M & F proper names, respectively, in Kruševo Aromanian (Gołąb 1984a: 81), where they do not take definite marking (the plural with a + -lor includes proper names). Moreover, Ohrid-Struga Frasheriote Aromanian has al and ali, respectively for all M/F def IDOs and has completely lost the gen-dat definite articles, thus bringing it closer to Macedonian in its lack of declension (Markovikj 2007: 75–77), e.g., Kruševo a vãsilelui, Frasheriote al caru ‘to the king.’ In Meglenoromanian, according to Atanasov 2002: 213–215, the dative case, in marking indirect objects, is always preceded by la or lă, which continue in some form a Vulgar Latin *illac-ad (Vrabie 2000: 48; cf. Caragiu-Marioțeanu 1968: 92). An example from Meglenoromanian is in (6.16), from Atanasov 2002: 214:
(6.16)
dă-lă la cǫ́ń di măncári give.impv-to.them to dog.pl some food ‘Give the dogs some food!’
Romanian also has the prepositional construction with la as a stylistic variant, so that (6.17), from Mihail 2013: 150, is a possible variation on (6.15b):
(6.17)
Ion dă mâncare la animale. Ion give.3sg food to animal.acc.pl ‘Ion feeds the animals.’ (lit., ‘Ion gives food to the-animals’)
Early Modern Romanian (sixteenth to eighteenth centuries, see Chapter 1, Footnote footnote 59) also had variation, with bare dative case (Mihail 2013: 149) but also prepositional phrases headed by a (Salvi 2011: 342) possible for the marking of indirect objects, as in (6.18a) and (6.18b), respectively:
a.
Să se atingă stricatului (Coresi, Tetraevanghelul) dms refl touch.sbjv.3sg depraved.def.dat ‘to (put the) touch (to) the depraved’ b.
o děde el a lucratori (Coresi, Tetraevanghelul) it.acc gave.3sghe to worker.pl ‘He gave it to the workers.’
Thus the replacement of earlier datives by prepositional phrases in Balkan Romance would seem to have its roots in Early Latin and could well be, therefore, independent of the development seen in Greek. Still, the emergence of the prepositional possibility as the sole marking in Aromanian and Meglenoromanian, in contrast to the variation still seen in Romanian, is suggestive of a contact explanation, with the loss of the gen-dat article in Ohrid-Struga Frasheriote almost certainly being associated with contact with Macedonian.
As for Slavic, the situation in OCS is the starting point for any consideration of developments with cases. In OCS, the dative case, by itself, was the primary means of marking indirect objects, occurring with “verbs of giving, saying, promising, commanding, scolding, rebuking, annoying, pleasing, liking, believing, serving, helping” (Lunt 2001: 148–151). Nonetheless, a prepositional alternative to the bare dative case can be found, and ultimately that is what prevails in Balkan Slavic; both Bulgarian and Macedonian mark an indirect object with na, which originally had locational (‘on,’ with locative case) and directional (‘onto, to, toward,’ with accusative case) senses. In Bulgarian and eastern Macedonian, this marking has even extended to the 1pl/2pl pronouns, e.g., na nas vs. nam ‘us.dat.pl.’ In Bulgarian, even the 1pl/2pl clitics have merged: ni/vi dat and acc, vs. Macedonian acc ne/ve but dat ni/vi. Marking the pronominal dative with na is expanding to the 1sg/2sg, e.g., modern Bulgarian and nonstandard colloquial Macedonian daj mi na mene ‘give me.dat to me.obl’ vs. standard daj mi mene. This is thus a change in the direction of analytism that is taking place in real time. While the fact that this phenomenon is already present in the eastern dialects is undoubtedly a factor, the influence of Albanian, where the pronouns in the first two persons do not distinguish accusative and dative, could also be a contributing factor (although Albanian does have a distinct ablative for strong pronouns).
The prepositional alternative in question in OCS is the preposition kъ ‘to, toward,’ which governs a dative nominal, being used with verbs of saying, but both the status of this construction and its importance for understanding the emergence of the na construction are a matter of some debate. Lunt 2001: 151 argues that the use of kъ for indirect objects “is a clear case of Greek influence: almost invariably when the Greek has the dative case, so does OCS, but when Greek has πρός + acc (a normal construction), OCS has kъ + dat.” He concludes, “There is no evidence that this usage was ever a part of a spoken dialect.” O. Thomason 2006: 303, 328, however, focusing on the instances where kъ does not match Greek πρός, considers the addressee function in OCS to be expressed in “several instances” (nine in Matthew, eight in Luke) through spontaneous prepositional constructions. And prior to her, Rusek 1964: 118–119 found that in the later (thirteenth century) Hludov Triodion, while the OCS pattern is evident by which prepositional constructions with the verb rešti ‘say’ mostly follow the Greek, nonetheless there were spontaneous instances of the prepositional construction, and in greater numbers than in OCS. Such a situation is what gave Meyer 1920 reason to conjecture that the emergence of a prepositional construction with kъ was what made possible the later construction with na (+ accusative). Meyer’s view is similar to what is suggested above with regard to early tendencies towards prepositional usage in Classical Greek and Latin. Nonetheless, Wahlström 2015: 64 doubts Meyer, giving the quite reasonable argument that “the more frequent spontaneous use of the prepositional construction could be indicative of the phrase having become a formulaic expression, whereas in OCS the construction resulted more often from word-for-word translations.”
Thus the real question is how and when the na alternative to the dative arose. Wahlström 2015: 82–83 gives a convenient summary of the facts and of various interpretations of them:
Meyer [1920], Rusek [1964], and Duridanov [1956] believe that the texts exhibit an already relatively well-developed analytical structure, homonymic to the dative case, in the spoken language of the 13th century. On the other hand, Mirčev [1957, 1978] and Lunt [1965], for example, claim that the manuscripts do not offer proof to support the early emergence of the na-construction nearly as clearly as the others suggest, although they do not categorically deny this possibility. … Nevertheless, while not yet emerging as a replacement for the dative, the preposition na with its increased use in abstract senses in the M[iddle]B[alkan]-S[lavic] corpus is indicative of its semantic bleaching, a necessary condition for its more grammatical role.
Thus, it would seem that there is simply not enough of a purely language-internal nature to offer a compelling account of the rise of na in Balkan Slavic as a replacement for the dative. This suggests that some recognition of the role of language contact is needed.
With regard to reaching a full understanding of the convergence here, there are several key facts. First, there is the presence in each of the languages of the seeds of a tendency towards the prepositional strategy for indirect object marking. Second, it seems that no one language had enough internal wherewithal, in terms of internal forces at work, to cause the prepositional alternative to prevail. To these considerations can be added the fact that a parallel in the marking of direction and the marking of indirect object is not uncommon cross-linguistically (cf. English to, the Turkish dative case, the Finnish allative case, inter alia). Putting all of these strands together, we are led to a situation in which multilateral contact leading to mutual accommodation must be considered. That is, no one language triggered the change, but contact among the different languages brought out and enhanced each language’s tendency towards and potential for analytism with the prepositional strategy. The result was that they all moved in the same direction, steered that way essentially by parallel selection from a pool of variants in a mutually reinforcing way. This sort of model seems appropriate for other convergent features in the Balkans as well; see §7.7.2.1.4 regarding its application to the developments with the infinitive, for instance. In this way, it gains plausibility by its utility as a more general basis for an explanatory account.
6.1.1.1.3.3 Other Developments with Accusatives and Datives
Besides the specifics of interplay between accusatives and datives, or in some instances, dative successors, in the marking of indirect objects, there are some developments with these cases that are worthy of mention for how they do, or do not, relate to other developments in the Balkans.
First, as noted in §6.1.1.1.3.1 and developed more fully in §7.8.2.2.6, Romani developed a possessive construction with an accusative marking the possessor (and the possessum too, as it happens), illustrated in (6.19):
(6.19)
(man) si ma(n) duj čhave (me.acc) is me.acc two children ‘I have two children.’Footnote 45
The accusative possessor here has taken the place of the early Indic genitive, which itself replaced a presumably inherited dative of possession. It is an inheritance from Middle Indic that is unparalleled in the Balkans and thus is noteworthy for its divergence from the Balkan norm; still, given that the Romani case system differs significantly from what is seen in other Balkan languages, this divergence is not surprising. As also noted in §6.1.1.1.3.1, Romani also has accusative for indirect object with ‘give,’ e.g., de man! ‘give me [it]’ (but, e.g., phengjum tumenge ‘I told y’all’ (lit., ‘tell.1sg.aor you.pl.dat’)’ [Jusuf 1974]).
Second, in Greek, the genitive case, which assumed many functions of a declining dative in earlier times (see §6.1.1.2 for details), has itself been declining in use in the past few centuries, especially in the plural. This development is quite apart from, or perhaps better, in addition to, the fact that indirect object marking was taken over by the accusative in the north, as discussed in §6.1.1.1.3.1. This decline was noted as early as Thumb 1895.Footnote 46 There are now individual nouns, e.g., κότα ‘hen,’ and whole classes of nouns, e.g., diminutives in -άκι, that lack a genitive plural altogether. The retreat of the genitive plural has a variety of causes, all conspiring, as it were; the most recent and most thorough examination of it in Standard Modern Greek, Sims 2015: 206, states that it “likely arose from multiple factors, including lack of paradigm cohesion [e.g., regarding stress placement in certain inflectional classesFootnote 47], the easy availability of a prepositional phrase as a way to avoid the genitive plural [e.g., in the marking of indirect objects], and the fact that genitives [sic] are associated with Katharevousa and prepositional phrases with dhimotiki.” The spread of prepositional alternatives is interesting, since some 100 years earlier even, Thumb 1912: §44.2 noted that “In North. G[r]k. dialects (e.g., Thessaly, [Greek] Macedonia) the gen[itive] has all but disappeared … i.e., the prep[osition] από has largely ousted it,” an observation he repeats in §161.6.1, where he specifically calls this a “way the gen. pl. can be avoided.” There is thus a purely language-internal raison d’être to the retreat of the genitive. However, the fact that it is most thoroughly realized in northern dialects, in an area in which there has been sustained contact between Greek and Slavic, makes it likely that there is Slavic involvement in the way this development has been extended in the north. This is especially so since the use of από that Thumb refers to mirrors that of western Macedonian od ‘from’ and eastern Macedonian and Bulgarian na, in the marking of possession, partitivity, and other originally genitive functions.
6.1.1.2 Genitive/Dative Developments
Although several developments of Balkanological interest center on genitives and datives and their typical functions, there is one, the merger of genitive and dative cases, that has attracted considerable attention over the decades. In fact, this particular case-merger is one of the oldest-cited Balkanisms, being mentioned, for instance, in Miklosich 1862 (with an extensive treatment also in Sandfeld 1930: 185–191 along with other related genitive/dative developments). The basic observation concerning the genitive and the dative is that from earlier states in which these cases were robustly distinct, Albanian, Balkan Slavic, and Greek, as well as most of Balkan Romance, have developed to a stage with no formal (i.e., surface) distinction between the shape of the genitive case and the shape of the dative case in either the singular or the plural.Footnote 48 However, while the fact of merger is indeed shared among the languages and there are striking similarities to note, the paths by which it was achieved were somewhat different for the different languages, and details of its realization differ, along formal, functional, and chronological dimensions. Moreover, it is necessary to distinguish between mergers in form and mergers in function, and in some instances to consider nouns and pronouns separately, for the mergers are not uniform across all of these levels and nominal types. The formal facts are considered first, with the chronological details following. Functional developments are taken up in later sections, with some treated in considerable detail. Following the discussion of the functional side of the case-merger, functional aspects of the genitive and dative cases that are not tied to the merging of the cases are presented as well.
6.1.1.2.1 Language-Particular Formal Developments
As noted, each language shows its own particular manifestation of the merger of genitive and dative cases. Some generalizations are possible across some of the languages, but for the most part, the individual languages must be taken on their own terms.
6.1.1.2.1.1 Greek
In Greek, for both nouns and pronouns, the genitive form replaced the dative form, in both the singular and plural, although the more usual way of describing this development in the literature is in terms of the loss of the dative.Footnote 49 As shown in Table 6.5 with the noun ‘man,’ the first person pronouns, and a demonstrative pronoun, the earlier genitive/dative distinction gave way to just a single form in later Greek continuing the earlier genitive forms and at least some of the functions of both earlier forms (see Table 6.5).Footnote 50
| gen.sg | dat.sg | gen.pl | dat.pl | ||
|---|---|---|---|---|---|
| AGrk | ἀνθρώπου | ἀνθρώπῳ | ἀνθρώπων | ἀνθρώποις | ‘man/men’ |
| ἐμοῦ | ἐμοί | ἡμῶν | ἡμῖν | ‘I/we’ | |
| αὐτοῦ | αὐτῷ | αὐτῶν | αὐτοῖς | ‘this/these’ | |
| ModGrk | ανθρώπου | –– | ανθρώπων | –– | |
| εμένα/μου | –– | εμάς/μας | –– | ||
| αυτού | αυτών | ||||
| (αυτουνού)/του | –– | (αυτωνών)/τους | –– | ||
| (earlier, των)Footnote * |
* Greek forms after “/” are weak forms; those in parentheses are independent possessives.
Moreover, as discussed in §6.1.1.1.3.1, there are further developments with the genitive in its indirect object function, in that accusatives in northern Greek dialects for nouns and pronouns, and in southern dialects (and thus the standard modern language) for pronouns, extend their domain at the expense of indirect object genitives. This represents a way in which the genitive/dative merger is undone, so to speak, in that the genitive/dative in one guise, namely that of indirect object, separates off from the genitive/dative in a different guise, namely that of possession. To the extent that indirect object is a prototypical dative function and possession a prototypical genitive function, this secondary development in northern Greek dialects represents a reconstitution, in a different form, of a distinction between the genitive and the dative, contrary to the general merger of such cases. A similar outcome involving prepositions and the marking of possession is discussed in §6.1.1.2.4.2.
6.1.1.2.1.2 Balkan Slavic
In contrast to Greek, in Balkan Slavic there was generally a replacement of the genitive by the dative, but the genitive replaced the animate masculine (o-stem) and pronominal accusative – already attested, albeit not with complete consistency, in OCS – and eventually all or most accusative pronouns.Footnote 51 Thus the old genitive pronoun survives but with a new function, as direct object. With regard to cases in general, in OCS there were distinct dative and genitive forms throughout the noun declension (except singular i-stems). Later, though, in the period post-OCS proper, the preposition na emerged as an analytic competitor for dative marking (expanding from an earlier accusative-locatival function for CoSl na ‘at, on, onto’; from that, na took on the value of genitive, extending the analytic structure to another case.
But even in parts of Balkan Slavic, such as the Macedonian of the Kostur (Grk Kastoria) region, where definite nouns still show distinct dative forms, built with an adjoined inflected dative definite article, e.g., pop ‘priest,’ def popo, dat.def popotomu (see §6.1.1), the function is that of possession as seen in all the examples given by Šklifov 1973, e.g., Popotomu sin se ženi ‘The son of the priest gets married.’ Thus, while the form contains a survival of the dative, the usage is equivalent to an old genitive or a possessive adjective. This means that ultimately, as far as form was concerned, it was the dative, or a prepositional dative surrogate, that won out over the genitive in Balkan Slavic. (See Mazon 1936: 54ff. on the Macedonian spoken around Korča (Korça) and elsewhere in Balkan Slavic.)
6.1.1.2.1.3 Balkan Romance
In Balkan Romance, in a development that is evident only in masculine definite forms, since genitive and dative in ā-stem (first declension) singular nouns had merged already in Latin (e.g., aquae ‘of water; to/for water’), the dative and genitive merge, with the dative replacing the genitive; this development, however, occurred only in the singular of nouns; in the plural, the opposite occurred, with the genitive form prevailing. That is, in a feminine noun like ‘water,’ the Romanian genitive/dative ape ‘of/to water’ derives its ending from the merged genitive/dative -ae of Latin (aquae), whereas the nominative/accusative apă has its ending -ă from the respective Latin endings -a/-am. Masculine nouns have a genitive/dative in the singular in -u-lui (definite, e.g., elev-u-lui ‘of/to the pupil’) that is from Latin illui ‘to/for this’ (where the ending -ui is analogical based on cui ‘to/for whom’) and/or a Vulgar Latin form illaei, both dative forms. The plural genitive/dative, by contrast, has the ending -lor (e.g., elev-i-lor ‘of/to the pupils’), from the Latin genitive plural illoru(m) ‘of them; their’ (versus dative plural illīs). Thus different cases were generalized in the singular as opposed to the plural of nouns, though the result in each instance was a merger of genitive and dative.
Still, despite the general fact of this overall merger, some differences in its realization in the various languages can be discerned. In particular, Meglenoromanian, as described by Capidan 1925a: 147, has genitive forms that systematically differ from dative forms as to the preposed particle they occur with (genitive lu versus dative la)Footnote 52 and for some nouns whether or not they require a postposed definite article (e.g., singular -u, plural -il or -li for masculine nouns); for instance, for locu ‘place,’ Capidan gives the following forms (see Table 6.6).
| sg | pl | |
|---|---|---|
| nom | locu | locurli |
| gen | lu locu, locului | lu locur, (-li) |
| dat | la loc(u) | la locurli |
| acc | locu | locurli |
| voc | loculi | locurli |
It might well be an analytical question, a matter of interpretation, whether these differences count as a “formal” distinction in the cases, since they hinge on the appearance or not of certain particles. However, it is clear that the situation Capidan describes is different from, say, that found in Romanian, where there is formal syncretism between genitive and dative in all genders and numbers in nouns, as in the examples given above, and pronouns (e.g., lui means both ‘to him’ and ‘his,’ lor is both ‘to them’ and ‘their’).
Moreover, the situation is somewhat different as far as pronouns are concerned. For Meglenoromanian, the pronominal dative and genitive cases are mostly identical, differing only as to the preposed lu versus la, but Capidan 1925a: 154 gives the interrogative pronoun/adjective cári ‘which?, who?’ as having simply la cari for the dative but a range of variants for the genitive: curui / la curui / cari / la cari. According to Atanasov 1990: 203, this same interrogative can marginally make a formal distinction between the genitive c(ă)rúi and the dative c(ă)rúi in that the genitive can occur without a preceding particle (lu, ăl, or ău) while the dative never occurs without a particle (la or lă). This distinction has continued more recently (Atanasov 2002: 221), though only in the speech of the oldest generation: cari ‘who’ pe cari ‘who.acc,’ la/lă cari ‘who.dat’ but lu/ăl/ău cruj ‘of.whom, whose (gen).’ Elsewhere, however, the dative and accusative are distinct, and the genitive is identical to the dative.
For Aromanian, according to Capidan 1932, the genitive and dative have merged for all nouns and all pronouns. It is thus just like Romanian in this regard.
Putting all these facts together, one can say that the trajectory of the genitive-dative merger is different across Balkan Romance. The situation in Aromanian differs somewhat from that in Meglenoromanian, and is more like what is seen in Romanian. Thus while there is general uniformity across Balkan Romance on this feature, there are differences that point to the languages carrying out the merger each in its own way (cf. Maiden et al. 2021: 79ff.).
6.1.1.2.1.4 Albanian
The genitive/dative merger in Albanian is found in all noun classes and inflectional categories, singular and plural, definite and indefinite; representative feminine and masculine forms are as follows (see Table 6.7).Footnote 53
| ‘girl’ | nom.sg.indf | vajzë | nom.sg.def | vajza | nom.pl.indf | vajzë | nom.pl.def | vajzët |
| gen.sg.indf | vajze | gen.sg.def | vajzës | gen.pl.indf | vajzëve | gen.pl.def | vajzëve(t) | |
| dat.sg.indf | vajze | dat.sg.def | vajzës | dat.pl.indf | vajzëve | dat.pl.def | vajzëve(t) | |
| ‘city’ | nom.sg.indf | qytet | nom.sg.def | qyteti | nom.pl.indf | qytete | nom.pl.def | qytetet |
| gen.sg.indf | qyteti | gen.sg.def | qytetit | gen.pl.indf | qyteteve | gen.pl.def | qyteteve(t) | |
| dat.sg.indf | qyteti | dat.sg.def | qytetit | dat.pl.indf | qyteteve | dat.pl.def | qyteteve(t) |
Despite this formal merger, there is differentiation between genitive and dative at a level other than case marking per se. In particular, the dative is used as the (indirect) object of a verb, and is always accompanied by object reduplication on the verb in normative usage (as also in Macedonian; see §7.5.1); the genitive, by contrast, is always preceded by a particle of concord. Thus the cases are distinguished morphosyntactically.
Interestingly, under one interpretation of the prehistory of the Albanian case endings, the same singular versus plural differential seen in Romanian can be observed. That is, the masculine indefinite singular genitive/dative ending -i seems to derive from an earlier dative ending, PIE *-ōi (< *-o-ei, Sh. Demiraj 1985: 275–276, 1054) or maybe just from *-ei. The source of the plural -ve is less certain; among the possibilities are the preceding, as listed by Sh. Demiraj 1985: 283–284, 1055, with mention of the scholars who originated the particular etymology, but ultimately rejected it (Table 6.8).
| -ve < *-bhyas (Meyer; cf. Skt dat/abl.pl -bhyas) |
| -ve < *ovōm, genitive plural of demonstrative ovŭ (Skok) |
| -ve < *-ois (Durante; cf. Grk dat.pl -οις, Skt inst.pl -āis) |
Demiraj also rejects Bokshi’s suggestion that the plural showed the same development as that in the singular.
On the phonological side of these possibilities, deriving Albanian e from earlier *ō is quite well supported (e.g., ne ‘us.acc’ from PIE *nōs; cf. also tërmet ‘earthquake’ borrowed from Latin terra mōta), while the other hypotheses require special assumptions phonologically. Ultimately, Demiraj opts for a derivation from *-ōm, as in the Vedic genitive plural suffix -ām or Latin (third declension) -um; the source of -v- is uncertain but it could very well be a hiatus breaker that was phonetically triggered when *-ōm was added directly to a theme vowel – note that *-ō- as a round vowel is conducive to the emergence of an on-gliding labial sound like -v- (if originally [w]).
Pronouns also show the merger completely. This is particularly evident with the demonstrative pronouns as there are special possessive pronouns that show their own particular behavior; thus for ky ‘this(m),’ the form këtij means ‘to/for this’ (dative) and ‘of this’ (genitive), and for ato ‘these(f),’ the form atyre F/M means ‘to/for these’ (dative) and ‘of these’ (genitive) (the genitives, as noted above, always occur with a particle of concord).
6.1.1.2.1.5 Romani and Turkish
Genitive-Dative merger in Albanian, Balkan Romance, Balkan Slavic, and Greek was already underway by the middle ages, i.e., prior to or at the time of the arrival of Romani and prior to the arrival of Turkish (leaving to one side Kipchak and other Turkic, see §1.2.1.11), and neither Romani nor Turkish shows much in the way of such case syncretism. Balkan Turkish maintains an inherited genitive-vs.-dative distinction. Similarly, Balkan Romani for the most part also keeps an inherited distinction between these two cases, though the actual realization of the case markers in question is different from earlier stages of Indic, being formed by what are historically postpositions instead of case-endings proper (although they are now, synchronically, case endings; see Friedman 1991). The basis for case formation in Romani is the oblique stem, normally -Vs for masculines, -(j)a for feminines, and -Vn for the plural. Romani has a distinction between thematic or oikoclitic and athematic or xenoclitic stems. The former are native words and pre-Byzantine loanwords that have the stem vowel -e- for V and the latter are post-Byzantine loans with some other vowel, e.g., dženó ‘person’/obl.sg dženés-, džépo ‘pocket’/obl.sg džepós-.Footnote 54 Table 6.9 is illustrative of basic masculine and feminine declensions.Footnote 55 Note that the forms given for the genitive are for a masculine nominative singular head. The vowel(s) in italics are i for a f.nom.sg head and e elsewhere, e.g., Esmakoro dad ‘Esma’s father,’ Esmakiri gili ‘Esma’s song,’ Esmakere gilja ‘Esma’s songs,’ etc.
| Singular | Plural | |||
|---|---|---|---|---|
| m | f | m | f | |
| Nominative | čhavo | čhaj | čhave | čhaja |
| Vocative | čhavea | čhaje | čhavale[n] | čhajale[n] |
| Accusative | čhave[s] | čhaja | čhaven | čhajen |
| Dative | čhaveske | čhajake | čhavenge | čhajenge |
| Locative | čhaveste | čhajate | čhavende | čhajende |
| Ablative | čhavestar | čhajatar | čhavendar | čhajendar |
| Instrumental | čhave[s]a<r> | čhaja[s]a<r> | čhavensa(r) | čhajensa(r) |
| Genitive | čhavesk[or]o | čhajak[or]o | čhaveng[or]o | čhajeng[or]o |
In those Romani dialects where [Vr] is lost, there will sometimes be homonymy between dative and oblique-agreeing genitives.Footnote 56 A number of Balkan dialects have merged the dative and locative, favoring one or the other (both in terms of the case suffixes and the related prepositions t-, and k). Also the prepositional genitive marked by a preposition of ablative origin [ko]taro ‘from’ (cf. Macedonian od, Bulgarian ot) is gaining ground (see Igla 1999).
6.1.1.2.2 Chronological Considerations
The evidence of direct attestation for some of the languages along with the cross-linguistic material in the preceding section (§6.1.1.2.1) allows for some definitive conclusions to be drawn about the dating of the genitive/dative merger. Determining the timing of the merger for Albanian, however, is difficult, other than observing that it must surely be early and predate the split between Geg and Tosk due to the fact that it is complete in both major dialects. And, Sh. Demiraj 2002: 488 claims that “they owe their origin to a common case ending,” suggesting a very early, even Albanoid, syncretism of genitive and dative. As for the other languages, it emerges that each seems to show its own particular chronology for this merger.
The developments with the genitive and dative in Greek have their roots in the Hellenistic and Roman periods where, in texts as early as the Septuagint of the third to second century BCE, but also in authors such as Polybius (second century BCE), prepositional replacements for what would have been datives in Classical usage are evident (cf. Browning 1983: 36–37; Horrocks 2010: 97, 107–108, 184–186). For instance, in the first century BCE, in the papyri, first the accusative and then the genitive replace dative pronouns, with further spread of the genitive over the dative to substantives in the early centuries CE. Moreover, by the first century CE, the dative as object of the preposition ἐν ‘in’ had given way to another preposition, εἰς ‘in; to’ with an accusative object. More generally, a reduction in the use of the dative, compared to Classical Greek, is evident in New Testament Greek. Blass & Debrunner 1961: 100, while emphasizing that the dative is “still retained in the N[ew]T[estament] in a wide range of usages,” nonetheless observe that “the dative was exposed to a greater extent than either the accusative or genitive to the encroachment of various prepositions … on the function of the simple case.” Considerations such as these are why the Greek developments are more usually characterized as the loss of the dative rather than merger with the genitive.
Still, Horrocks 2010: 185, drawing on the important work of Humbert 1930 and Trapp 1965, observes that “Although it is clear that the domain of the dative was everywhere in retreat, especially with regard to its old adverbial functions (time, place, instrument etc.), Humbert 1930, the only fully detailed study of the decline of the dative to date, argues that the dative in its core grammatical function of marking a secondary object did not finally disappear from the “spoken language of native speakers until the 10th century AD.” Moreover, he sees the dative as “a fixture of the written Koine throughout the Byzantine period and beyond” though he notes that its grammatical function was “steadily transferred … to the genitive: for indirect objects (both clitic pronouns and full noun phrases)” (p. 284).
For Balkan Romance, the merger appears to have its beginnings in the merger seen in Latin of the feminine singular dative and genitive, e.g., aquae ‘to/for water; of water,’ a change which happened very early in the history of Latin.Footnote 57 This merger apparently progressed in later Latin – Stan 2013a: 263, citing Iliescu 2008: 3268, observes that “Romanian appears to have preserved the syncretism of the genitive with the dative from Late Latin,” a position adopted by others as well (e.g., Adams 2013: 319). By the time of attested Romanian in the sixteenth century, the merger was apparently complete, but the fact that the merger was not yet complete in Meglenoromanian in 1925 and even more recently suggests that the full generalization of genitive/dative syncretism in Romanian, and Aromanian too, for that matter, as discussed in §6.1.1.2.1.3, came after the emergence of the individual languages.
As for Balkan Slavic, it must be noted first that genitive and dative are distinct in Old Church Slavonic in all noun classes and all numbers, except for the singular of i-stem nouns (e.g., nominative kostь ‘bone,’ with genitive/dative kosti, as opposed to, e.g., o-stem nominative rabъ ‘slave’ with genitive raba and dative rabu, or u-stem synъ ‘son’ with genitive synu and dative synovi). Given that the vast preponderance of stem classes, therefore, distinguished these cases, it is most likely that the Balkan Slavic merger began post-OCS.Footnote 58 The exact pathway by which this merger occurred is undoubtedly a matter of the receding of the genitive in its original uses, since, as indicated in §6.1.1.2.1.2, genitive shifted to being a marker of direct objects, usurping a key function of the traditional accusative, and datives are the basis for the merged genitive/dative case form. Wahlström 2015, the most recent treatment of the phenomenon of case-loss in Balkan Slavic, posits (p. 82, Table 4), for what he calls Middle Balkan Slavic, the Slavic of the thirteenth to sixteenth centuries in the Balkans, a case system with merged genitive and dative, e.g., for ‘son’ (see Table 6.10).
| sg | pl | |
|---|---|---|
| Nominative | sin | sinov[e/i] |
| Oblique | sina | sinov[e/i] |
| gen/dat | sinu | (sinov-[om/em]) |
The most important subsequent, or probably, at least partly contemporaneous development was the replacement of genitives and datives by prepositional expressions headed by na, which when taking an accusative complement originally indicated goal of motion, i.e., ‘to.’
Understanding the pathway for that particular line of development requires attention to the functional side of the merger; this has been hinted at in passing up to here but taken up in earnest in the sections that follow.
The upshot of the chronological considerations discussed here is that the medieval period would seem to be the temporal locus for the generalization of a merged genitive/dative case in Balkan Slavic and Greek, and perhaps most of Balkan Romance, though it has yet to go to completion in Meglenoromanian. On the other hand, Greek and Balkan Romance show early signs of a case merger, and Albanian probably is more in line with them, chronologically, based on Demiraj’s assessment. It is difficult, therefore, to pinpoint an era when all of the languages were shifting, though importantly, all show movement in the same direction across their attested or reconstructible history, including periods of contact of one or more languages with others.
6.1.1.2.3 Functional Considerations: Indirect Object
From a functional standpoint, a merger of dative and genitive cases presents some interesting points to consider. The dative case, in languages that have such a form, is widely seen as having the marking of indirect object as a, if not the, primary function,Footnote 59 though other functions of the dative, especially marking benefactives or malefactives, play a role in shedding light on the merger. Genitive case, by contrast, is generally seen as being primarily a relational case, marking especially possession. Thus a merger required grammatical adjustments in each of the languages, depending on the direction of the merger. The continuation of the marking for these primary functions is thus an issue to attend to in tracing the history of the genitive-dative merger in the various Balkan languages.
In Greek, the prototypical dative function of marking the indirect object was absorbed by the genitive but also handed off to prepositions, most notably εἰς (later form: σε) ‘to’ governing an accusative complement. But the benefactive function is a key to understanding the merger. That is, it is well documented that interchanges between datives and genitives per se occur earliest in the personal pronouns (Humbert 1930; Stolk 2015; Wahlström 2015: 112). Moreover, prenominally positioned genitives could easily be interpreted as clausal second position benefactives, with no significant change of meaning; this second century CE example from a papyrus text (Humbert 1930: 171) illustrates the potential for reinterpretation:
(6.20)
ἀγόρασόν μου τὸ μέρος τοῦ ἐλεῶνος buy.impv me.gen the.acc.sg part.acc.sg the.gen.sg olive.grove.gen.sg ‘buy my part of the olive grove’ ~~> ‘buy the part to-my-benefit of the olive grove’
From such a potential overlap, it is claimed, the genitive could start to absorb some of the functions of the dative case, leading to a genitive being generalized at the expense of the dative. As Stolk 2015: 94 puts it, “From the role of the malefactive/benefactive the meaning of the genitive case could have been extended further into goal-oriented roles such as addressee and recipient, which are commonly denoted by the dative case in Ancient Greek.” The availability of prepositional alternatives to other, even possibly primary, dative functions would have played a role as well in the demise of the dative, and thus in the ascendancy of the genitive. Moreover, as noted in §6.1.1.2.2, the use of the dative as object of prepositions was also receding, and in Modern Greek now, except with prepositions of a more learnèd origin (borrowed from higher registers into colloquial use), accusative case is the norm for prepositional objects; the only colloquial exception is the use of weak genitive personal pronouns with the more adverbial (as opposed to relational) prepositions, e.g., μπροστά μου ‘in-front of-me,’ πίσω τους ‘behind of-them.’Footnote 60 And, there are instances of dative objects in Ancient Greek giving way to genitives in Modern Greek; as one such verb among several, Thumb 1910: §54 mentions ακλουθώ ‘to follow,’ whose Ancient Greek source, ἀκολουθέω, governed a dative.Footnote 61
The trajectory was the opposite in Balkan Slavic in that the ultimate form that the merger took, with the preposition na being used for dative and genitive functions, seems to have started with typical dative uses and moved to genitive uses. That is, based on its original meaning of goal-oriented ‘to,’ na as a marker of recipient (indirect object) would seem to reflect its marking of generalized movement towards something, with recipient being equated with the goal of motion. Cross-linguistic parallels such as the use of to in English or εἰς in Greek or Romance a (from Latin ad ‘to’) offer typological support to the view that this is a natural pathway of semantic and grammatical development. The oldest texts also show examples of dative constructions where the Greek has a genitive. Thus the Slavic pathway seems to have genitive functions ultimately absorbed by datives or dative surrogates (see §6.1.1.2.4 as to where possession fits in).
For Albanian and Balkan Romance, the functional pathway is not as clear. Since the genitive prevails in form in the plural whereas dative prevails in the singular, it may well be that both sorts of pathways, the genitive-absorbing-dative functions seen in Greek and the dative-absorbing-genitive functions seen in Balkan Slavic, could have been at work. There are several parallels within Romanian to the use of a dative-like element for possession. That is, it may be relevant that Romanian a – from Latin ad ‘to’ and thus most likely originally serving a dative function in marking indirect objects (as elsewhere in Romance) and still used in a dative-like fashion with some prepositions, e.g., contrar ‘contrary (to)’ or mulţumită ‘thanks (to)’– has come to be used, as Stan 2013a: 268 puts it, “as an analytic marker of the genitive relation” when inflectional case marking is impossible, as with a quantifier, as in mama a trei copii ‘mother of three children.’ Further, the Romanian preposition la (see Footnote §6.1.1.2.1.3, footnote 52), which is the usual way of marking indirect objects analytically in the present-day language, as noted by Stan 2013a: 269 “in informal language … is used to introduce referential genitive phrases,” as in mă-sa la fata asta ‘this girl’s mother’ (lit., ‘mother-her to girl.DEF this’). And, there is also the “grammaticalized functional element” (Stan 2013a: 265) al, probably derived from Latin ad with the definite article, which also marks dative and genitive functions in certain contexts, as in două cărţi ale professorului ‘two books.INDF of professor.DEF.DAT.’ Thus on several fronts, a dative-absorbing-genitive path of development can be presumed for at least parts of Balkan Romance for some elements, so generalizing that pathway to all such developments with the dative and genitive is not unreasonable.
Finally, for Albanian, as noted in §6.1.1.2.1.4, the cases are kept distinct morphosyntactically via object reduplication (for dative) and the use of the particle of concord (for genitive), even if there is formal merger of endings.
6.1.1.2.4 Functional Considerations: PossessionFootnote 62
The relevance of the marking of possession – the quintessential adnominal use of the genitive – for the merger of genitive and dative has already been touched on in preceding sections, but it bears further discussion. In particular, besides the issue of tracing pathways leading to the merger, it turns out that there are some clear contact-related aspects to other ways in which possession is expressed. Accordingly, the issues pertaining to the merger are laid out here first and then three subsections treat those other aspects.
6.1.1.2.4.1 Possession and the Genitive/Dative Merger
The key issue regarding the merger of genitive and dative cases is how datives could move into the marking of a typical genitive function. The discussion of Greek in §6.1.1.2.3 touched on one pathway from dative functions to genitive functions via the benefactive/malefactive use of weak object pronouns, with something happening to ‘my X’ being equivalent in a certain sense to something happening to ‘X’ to my benefit or disadvantage.Footnote 63
Another pathway is evident in Slavic. One major function for the genitive in adnominal use in OCS was indeed to mark possession (Lunt 2001: 146–147), e.g., duxъ otьca vašego ‘spirit father.gen your.gen’ (i.e., ‘the spirit of your father’), but as Lunt 2001: 149 observes, “The dative complement of substantives is semantically close to the possessive genitive.” He illustrates this with examples such as starěišiny ljudьmъ (Luke 19.47) ‘chiefs people.dat’ (i.e., ‘the chiefs of the people’). He goes on to say that “The possessive meaning is particularly common with the short dative personal pronouns mi, ti, and si: e.g., drugъ mi pride (L[uke] 6.6) ‘my friend has come’” [friend me.dat came.3sg.aor]. This “dative of possession” use makes sense from an Indo-European perspective, as it is found in other branches, especially Latin, where it “asserts the fact of possession” (Hale & Buck 1966: §374). Wahlström 2015: 49 has a telling observation here:
in the Codex Marianus the dative appears as the adnominal case in 24 percent of instances instead of the genitive.[FN: N = 363] However, the increase in the use of the adnominal dative in the M[iddle]B[alkan]S[lavic] manuscripts must be seen as reflecting the situation in the spoken language (see, e.g., Steinke 1968, 62–66). The dative gradually became the default adnominal case, as is evident in the dialects that preserve the dative: the genitive-accusative cannot be used to express possession, but the dative can be.
Regarding this last point, the discussion in §6.1.1.2.1.2 is pertinent here as it concerns dative forms collected by Šklifov 1973: 64 from the southwestern Macedonian of the Kostur (Grk Kastoria) region that have possessive meaning. And, while the genitive was clearly on the wane in terms of marking possession in Balkan Slavic, it did hang on, “occur[ring] sporadically in the vernacular-based Damascenes of the 17th century,” according to Wahlström 2015: 47, note 40.
Interestingly, that same pathway for a dative taking over the genitival function of possession is repeated in the way the preposition na enters the scene and moves from original goal-of-motion marking to recipient marking (i.e., indirect object – see §6.1.1.2.3) and ultimately to possession marking. This occurs even as na is taking over other dative uses found in earlier Slavic, most notably in marking benefactives (Wahlström 2015: 54), raising the possibility that a benefactive-to-possession pathway could have been open in Slavic too.
Ultimately, marking with na for genitive and dative functions prevails and datives, even with possessive function, are quite limited in terms of their dialectal range in Balkan Slavic. Thus, just as in Balkan Romance (see §6.1.1.2.3), a dative-absorbing-genitive path of development has several iterations in Balkan Slavic, with different material – dative case first and then with the prepositional dative surrogate (surrogates, in the case of Balkan Romance).
With regard to dative marking possession, it is worth noting here that Bulgarian, like Greek, favors clitic dative pronouns to mark possession, while Macedonian, like Balkan Romance and Albanian, favors pronominal adjectival constructions (cf. Topolińska 2010: 54). Romani likewise uses pronominal adjectives or genitives (which in Romani agree with their heads, like adjectives), although in North Macedonia, a prepositional construction using tar-o ‘from’ can calque the Macedonian use of od in this function (see below).
6.1.1.2.4.2 Other Means of Marking Possession
The discussion in §6.1.1.2.4.1 does not exhaust the marking of possession in Balkan Slavic and elsewhere in the Balkans. A further means found in Balkan Slavic is that for possession, and only possession, i.e., not for other functions of na, a different preposition occurs, specifically in western Macedonian. That preposition is od, which otherwise means ‘from’ – and continues to be used in that meaning as well – and it is used in place of na in that region to mark possession. The choice of preposition is typologically reasonable, as an affinity between ablative and genitive is found elsewhere within Indo-European; the Ancient Greek genitive, for instance, has absorbed the functions of the PIE ablative, which does not continue otherwise in Greek, and a similar development occurred between PIE and Balto-Slavic.
This prepositional marking of possession is of particular interest because a parallel shift occurs in certain dialects of Greek. In particular, in the northern dialect zone, the preposition από, which otherwise means ‘from’ (as well as ‘by,’ an option also found in Mac od, Alb prej, Aro di) has taken over most genitive functions. This development is noted by Thumb 1910: §§44.2, 161.6.1 for Thessaly and Greek Macedonia, giving examples such as σκέδιος από το σπίτι ‘plan of (lit., ‘from’) the house’ and has been verified recently by Chadzikyriakides & Spathas 2015, working mostly with younger speakers (aged between twenty-two and thirty-four); they give examples such as του σαμάρ απ του γαϊδούρ ‘the saddle of (lit., ‘from’) the donkey.’ Although there are Greek-internal factors possibly at play here, as emphasized by Mertyris 2015: 166, such as the loss of high vowels in northern dialects that would have deleted the masculine o-stem genitive -u in a large number of nouns, “language contact with Aromanian and western varieties of Makedonski [i.e., Macedonian, VAF/BDJ], where the ablative prepositions di and od are used respectively to mark possession, might be related to the highly advanced use of the preposition από as a possessive marker in Thessalian and Macedonian [Greek, VAF/BDJ] varieties.” It is not clear that Aromanian, which has and uses its genitive case, along with the preposition a, to mark possession, would have played much of a role here, though some uses of di correspond to genitives in related languages, e.g., sărbătoarea a curátil’i di Stă Marie ‘the-holiday of the-cleansing of (lit., ‘from’) St. Mary’ (Papahagi 1974: 467). Nonetheless, the geography of this development within Greek, occurring in just those regions where co-territorial languages that have at least some, and in the case of Macedonian, quite regular, use of a preposition that corresponds to Greek από, speaks strongly in favor of language contact having some effect. Moreover, the Macedonian Romani use of taro ‘from’ to mark possession is a calque on the western Macedonian usage that is increasing in frequency (cf. §6.1.1.2.1.5).
One aspect of the use of από in northern Greek and od in Macedonian for possession reveals an interesting way in which these locutions are counter-moves to the merger of genitive and dative in these languages. That is, από is not used for the indirect object function of the Greek genitive(/dative) nor is od used for the indirect object function of na in Macedonian. The same can be said for di in Aromanian or nga/prej in Albanian. This means that a distinction between genitive, in its prototypical possessive use, and dative, in its prototypical indirect object function, has been re-introduced in these dialects, since από and od are restricted to adnominal possessive use while other means are used for indicating indirect objects. A similar outcome with case marking and indirect objects is discussed in §6.1.1.2.1.1.
Going beyond the prepositional marking of possession in northern Greek and elsewhere, another means of marking possession, mentioned in §6.1.1.2.4.1 in connection with Old Church Slavonic, deserves attention, namely the use of weak dative pronouns or parallel forms. The OCS dative possessive pronouns are postposed (e.g., drugъ mi ‘friend to-me’ = ‘my friend’) and offer an alternative to the use of genitives or possessive adjectives, whether formed from pronominal bases, e.g., moj- ‘my,’ or from nouns, e.g., tektonov- ‘of-carpenter’ (tekton-).Footnote 64
The use of weak dative pronouns continues into later and modern Balkan Slavic, though with key differences between the languages.Footnote 65 In Bulgarian, they are used quite regularly with a wide range of ordinary nouns, so that knigata mi is the usual way of saying ‘my book’ (lit., ‘book.the to.me’). One noteworthy feature of these Bulgarian pronouns is that they always attach to the carrier of definiteness within the noun phrase, as discussed by Schürcks & Wunderlich 2003, so that they show mobility, as in (6.21):
a.
kniga-ta mi ‘book-def to.me’ = ‘my book’ b.
xubava-ta mi kniga ‘beautiful-def to.me book’ = ‘my beautiful book’ c. *xubava-ta kniga mi / *xubava mi kniga-ta
In Macedonian (and Aromanian), by contrast, these datives are used only with the closed class of kinship terms, e.g., tatko mi ‘my father,’ brat mi ‘my brother’; the equivalent of ‘my book’ would use a possessive adjective, mojata kniga (‘my-the book’), not necessarily with a definite article (Friedman 1993d). Interestingly, in this regard, Macedonian gives the impression of being more in line with other Slavic languages (cf. Russ moja kniga ‘my book’). On the other hand, it could be argued that Macedonian is closer to Albanian and Balkan Romance, while Bulgarian is closer to Greek (see below). Moreover, Macedonian has preserved the possessive adjectival formation more robustly than Macedonian (cf. Nicolova 2008: 119–120, and Koneski 1967: 119–120).
Greek has gone further than Bulgarian in that it now has only weak genitive pronounsFootnote 66 as the means of expressing possession, e.g., το βιβλίο μου/σου/του/της/μας/σας/τους ‘the book 1sg.gen/2sg.gen/3sg.m.gen/3sg.f.gen/1pl.gen/2pl.gen/3pl.gen’ (i.e., ‘my/your/his/her/our/your/their book’). They can always attach to the noun they modify but they can also move around within the noun phrase and attach to an adjective, e.g., το μεγάλο μου σπίτι ‘the big my.gen house’ (i.e., ‘my big house’); however, the attachment is not tied to definiteness as it is in Bulgarian.Footnote 67 This exclusive use of these pronominal forms is an innovation away from the Ancient Greek system where there were both possessive adjectives, e.g., ἐμός ‘my,’ ἡμέτερος ‘our,’ and genitive pronouns available to mark possession. The use of the possessive adjectives was declining severely in Hellenistic Greek. Blass & Debrunner 1961: 149 state that the “possessive adjectives … have to a great extent disappeared in the Hellenistic period and so also in the N[ew]T[estament].” Nonetheless, perhaps as learnedisms, the possessive adjectives can be found in Medieval Greek, e.g., in the fourteenth-/fifteenth-century romance Phlorios and Platzia Phlore, one can find ἐκ τὸ ἐμὸν τὸν πόθον ‘from the my the desire’ (i.e., ‘from my desire’), with the possessive adjective but just a few lines later ἀγάπη μου καλή ‘love of.me good’ (i.e., ‘my good love’) with the pronoun. It is telling, in terms of geography, that while the possessive adjectives remain in some Modern Greek dialects, this is so only in outlying, non-Balkan ones, specifically Pontic and Cappadocian, dialects of Asia Minor (Thumb 1910: §143.3).
A similar strategy is found in Balkan Romance, though of a more restricted character, and with greater options. The languages have both possessive adjectives, showing agreement with the nouns they modify, and under certain conditions, weak (enclitic) forms of those adjectives, and also the possibility of using dative weak object pronouns as “adnominal possessive clitics,” in the terminology of Nicolae 2013. Thus for Romanian, one finds, for instance for first person singular possession, the following (Nicolae 2013: 341–345), where (a) is the adjective, showing agreement, (b) is the enclitic adjective, also in an agreeing form, and (c) is the adnominal possessive clitic strategy with the dative pronoun:
a.
mâna mea ‘hand.f.sg.def my.f.sg’ = ‘my hand’ câinele său ‘dog.m.sg.def his.m.sg’ = ‘his dog’ b.
tac-su ‘father.m.sg.def his.m.sg’ = ‘his father’ c.
casa-mi ‘house.f.sg.def me.dat.wk = ‘house to-me’ = ‘my house’
Aromanian similarly has the three possibilities, as shown in (6.23), where a in (a) is a connective that occurs with the full form of the possessive adjectives:
a. ínima a mea ‘heart.f.sg.def to my.f.sg’ = ‘my heart’
b. soácra-ta ‘mother.in.law.f.sg.def-your.f.sg’ = ‘your mother-in-law’
c. cutsútlu-ts ‘knife.def-you.dat.wk’ = ‘knife to-you’ = ‘your knife’
The adnominal pronoun type of (6.23c) always adjoins to a definite form, either the noun, e.g., cutsutlu in (6.23c) (as opposed to indefinite cutsut) or a preceding adjective to which the definite marker has been affixed; these possibilities are shown in (6.24), from Nicolae 2013: 344:
a. cartea-i ‘book.def-3sg.dat.wk = ‘his/her book’ (lit., ‘book-the to-him/her’)
b. frumoasa-i carte ‘beautiful.def-3sg.dat.wk book’ = ‘his/her beautiful book’
Thus in Romanian, the adnominal pronoun is mobile within the noun phrase but tied to a definite element, as in Bulgarian. In Early Modern Romanian, there was a full set of forms for this type, plural as well as singular, and also a reflexive. However, in current usage, this type is obsolete, and its range is limited, in that there are only singular forms.
The similarities between Romanian and Aromanian in possessive possibilities suggest that these strategies reflect structures at least of Proto-Balkan Romance age. Most of the rest of Romance does not have the agreeing enclitic adjective type, though Salvi 2011: 337 documents it for southern and central dialects of Italian, so that it may reflect a Proto-Eastern Romance development. The fact that it is restricted to occurring with “kinship and social relation nouns” (Nicolae 2013: 341), a constraint found also with this type in Aromanian (Vrabie 2000: 51; Markovikj 2007: 73) – as in Macedonian – is suggestive of a common origin for the phenomenon in these two Balkan Romance languages.
Finally, in Albanian the situation arguably involves historically weak object pronouns to some extent. The language has pronominal adjectives that indicate possession and show agreement for the gender of the noun they modify. Their form in the third person, and in first and second person singular when modifying plural head nouns, is a combination of obligatory particle of concord with pronominal forms that match or closely parallel the corresponding genitive/dative case forms, and in the first and second person singular modifying a singular head and first and second plural the form includes an incorporated particle of concord that shows alternations in its initial consonant that reflect the incorporated particle of concord (e.g., t- in (6.25b) versus s- in (6.25d) mirroring what is found with a free particle of concord, and note i- in (a) and in (f)). Some examples may make this seemingly complex statement somewhat clearer:
a.
shoku im ‘friend.nom.m.def my.nom.m’ ‘my friend’ b.
shokut tim ‘friend.dat.m.def my.dat.m’ ‘to my friend’ c.
shoqja ime ‘friend.nom.f.def my.nom.f ‘my (female) friend’ d.
shoqjes sime ‘friend.dat.f.def my.dat.f ‘to my (female) friend’ e.
vëllait të tij ‘brother.dat.m.def pc.dat.m.def his’ ‘to his brother’ f.
emri i saj ‘name.nom.m.def pc.nom.m.def her’ ‘her name’ g.
emrave të tu ‘names.dat.m.def pc.dat.m.def your’ ‘to your names’ h.
vajza jonë ‘girl.nom.f.def our.nom.f’ ‘our daughter’
The possessive markers are usually post-nominal, like most modifiers in Albanian, but, as Newmark et al. 1982: 268 write: “singular first and second person adjectives may follow or precede” certain nouns, a set composed of kinship nouns and zot ‘lord, master,’ a noun of social relation; examples include im vëlla ‘my brother.nom,’ time motër ‘my sister.acc,’ sat motre ‘your sister.dat.’ Moreover, there is the construction, used exclusively with close kinship terms in the third person, that involves the particle of concord, e.g., i ati ‘his/her/their father.def.’
There are some parallels with possessives in the other languages, especially as to the mobility the Albanian forms show, as in Greek, Bulgarian, and Balkan Romance, and the involvement of kinship terms in conditioning some of the variation, as in Macedonian and Balkan Romance. However, there are also significant differences here, such as the ability of the Albanian elements to be in initial position in the noun phrase with certain nouns (e.g., im vëlla ‘my brother’), which is possible only with the full possessive adjectives in Balkan Slavic (e.g., mojata kniga ‘my book’). Also, the types of elements that show kinship-related restrictions differ across Balkan Romance, Macedonian, and Albanian.
Thus it is not clear if these phenomena can be connected. As for the Balkan Romance dative pronouns per se, some sources, e.g., Nicolae 2013: 343, disputes the Balkan character of this usage, since the behavior of the clitics is not uniform across all the languages.Footnote 68 Sandfeld 1930: 188, however, puts this forward as a Balkanism and some of the details of structure, history, and geography, e.g., the definiteness constraint on mobility in Bulgarian and Romanian, the fact of a period of intense Bulgarian-Romanian contact that could have allowed diffusion (in one direction or the other), and the Greek dialect distribution of the survival of possessive adjectives make at least contact involvement in the emergence of weak pronominal forms for possessive marking a compelling view; further consideration in §6.1.1.2.4.4 regarding a doubling structure adds strength to that view.
6.1.1.2.4.3 Reflexive Possession
One aspect of possession marking in the Balkans that is not directly tied to the genitive/dative developments but deserves attention nonetheless has to do with the overt marking of reflexive possession. Sandfeld 1930: 121–122 remarked on this and there is more to say.Footnote 69 While the facts in all of the languages are noteworthy, it is especially significant that Macedonian and Tosk Albanian show similar innovations in the direction of simplification that suggest a possible contact origin, given the location and timing.
Macedonian has inherited the Indo-European opposition between a reflexive root in *sw- (cf. Macedonian m.sg.adj svoj) and the other pronouns. In the rest of Slavic, the distinction between reflexive and nonreflexive third person possessive marking is strictly maintained. Thus, for example, in Russian, it is impossible to convey the ambiguity of the English sentence The directorx came with hisx/y wife; for his, Russian would have to choose between svoj- ‘his own’ and ego ‘his [someone else’s]’: Direktorx prišel so svojejx/s egoy ženoj. In English, the sentence could be disambiguated by adding own to his, i.e., The director came with his own wife, but such disambiguation would in fact imply that there was a reason to draw attention to the fact that the wife was not someone else’s.
Sandfeld 1930: 121–122 noted that in Macedonian the distinction between svoj/si and third person possessives (when coreferential with the subject) was not consistently maintained in nineteenth-century western Macedonian folklore texts, mostly from the Ohrid region, to which he had access. In fact, the strictness of the distinction between reflexive and third person has been lost, and as a result the Macedonian sentence equivalent to the “director” sentence above, Direktorot dojde so svojata/negovata žena, has exactly the nuances of the English ‘his own’ versus ‘his.’ That is, the pragmatic assumption with the third person possessive pronoun negovata is that it can refer to the subject if that is the normative relationship, and using the reflexive svojata in such a context suggests a nonnormative relationship. Thus, what is a grammatical distinction in Russian is a pragmatic distinction in Macedonian. By contrast, in Bulgarian, the reflexive/nonreflexive opposition (žena si / žena mu) is still strong among older speakers but is weak among younger speakers.
Thus, comparing the Bulgarian and Macedonian translations of the sentences under consideration, given here as (6.26ab) and (6.27ab), reveals that for the ordinary educated Bulgarian, overt reflexive (6.26a) is normal and overt nonreflexive (6.27a) is wrong or confusing if the possessive is to be co-referential with the subject; in Macedonian, however, as in English (and Greek), nonreflexive (6.27b) is normal, albeit ambiguous, whereas overt reflexive (6.26b) is overdetermined:
a.
Direktorăt dojde săs žena si (Blg) director.def came with wife refl.dat b.
Direktorot dojde so svojata žena (Mac) director.def came with own.def wife ‘The director came with his (own) wife’
a.
Direktorăt dojde săs žena mu (Blg) director.def came with wife him.dat b.
Direktorot dojde so negovata žena (Mac) director.def came with his.def wife ‘The Director came with his wife’Footnote 70
The disappearance of svoj from the Lower Vardar and Thrace, as well as Nestram (Grk Nestório) and Boboshtica (Mac Boboščica) to the west and much of the Rhodopes adjacent to Thrace, is suggestive of influence from Greek (see below on the Greek situation).Footnote 71
In Albanian, the element corresponding to Macedonian svoj is the possessive pronoun (m.nom.sg) i vetë.Footnote 72 In Tosk Albanian, the reflexive possessive has been completely lost, and thus in the formation of Standard Albanian, i vetë was introduced from Geg. For the “director” sentence given above, Geg can distinguish reflexive and nonreflexive: Drejtorix erdhi me gruen e vetax /e tijay, but Tosk has only … me gruan e tija. Moreover, this loss is characteristic of Arbëresh and Arvanitika as well as most (but not all) of the transitional dialects (Gjinari 2007: Map 94).
In Romanian (Graur et al. 1966: 157), the distinction between nonreflexive possessive lui and reflexive său that was realized consistently in older texts has been lost, replaced by a system in which the l-form is normal and the s-form is used only for disambiguation (Graur et al. 1966: 157–158). The situation is thus exactly the same as in Macedonian (and English). Based on Graur’s formulation and the dating of these texts (seventeenth century for the oldest), this change can be said to have been effected somewhat recently. Today in Romanian (Graur et al. 1966: 158), lui is felt as more colloquial and său as more bookish or polite.
In Aromanian, according to Cuvata 2009: 471, the reflexive s-form has been lost almost completely; Gołąb 1984a: 80, 88, however, reports for Krushevo traces of the originally reflexive s- possessive in a slightly different use with singular nuclear family kinship terms (tatã, afendu ‘father,’ dadã ‘mother,’ hilj ‘son,’ hilji ‘daughter,’ bãrbat ‘husband,’ muljer ‘wife’), e.g., a h’ili sai ‘to her daughter,’ strictly contrasting with the use of the original nonreflexive with other nouns, e.g., lúkurlu al’ei ‘her work.’ Meglenoromanian, however, according to Atanasov 2002: 219, has remnants of the s- possessive postposed after kinship terms (Meglenoromanian possessives normally precede the possessed). The Aromanian and Meglenoromanian preservation of the reflexive possessive with kinship terms is reminiscent of western Macedonian usage, where the use of dative possessive clitics is characteristic specifically of kinship terms (see §6.1.1.2.4.2).
Greek seems to be on a different plane from the other languages in this regard. It replaced the inherited reflexive possessive (that had a *sw- element in the third person) with an adjectival derivative of ίδιος ‘same; -self,’ namely δικός ‘own’ (with a possessive pronoun μου/σου/του, thus ‘my own; your own, etc.’). Greek is like English in terms of reference in the sentence in question from above; it could in principle disambiguate reference via με τη δική του γυναίκα ‘with his own wife’ (lit., ‘with the own his wife’) as opposed to με τη γυναίκα του, but the latter could also have a reflexive possessive reading in the right context.
Romani seems to have been quite tenacious in preserving some sort of autonomous reflexive marking, with various p- forms (cf. Middle Indic appan-, based ultimately on Old Indic ātman- ‘self,’ an Indic innovation) and some innovations appear to post-date the dispersal from the Balkans. More recent influence from contact languages also occurs. Thus, for example, in Macedonian Romani (Arli) the sentence Direktori avilo leske/pe romnjaja ‘The director came with his/his-own (pe) wife’ has the same contrasts as in Macedonian.
While the tendency to eliminate reflexive possessives can also be seen outside the Balkans, the geographical distribution of the elimination in the Balkans is striking. In the preservation of a reflexive possessive, Balkan Slavic comes across as the most innovating and Balkan Romance as the most conservative. Interestingly enough, Albanian is split along its oldest dialectal division in this respect as well, with Geg as conservative and Tosk as innovating. It appears that Romanian preserved the overall distinction longer than Aromanian and Meglenoromanian taken together, but they have preserved remnants in precisely that category where in Macedonian dative possessive pronouns are most likely to be preserved, i.e., close kinship terms. The overall elimination of distinctive reflexive possessive adjectives seems to be especially prevalent in the western and southern regions, in areas in contact with Greek, i.e., Aegean Macedonia and Thrace, as well as western Bulgarian and adjacent Torlak BCMS dialects. The weakening of the possessive pronominal use of (i) vet in Albanian occurs precisely where contacts with Macedonian, Aromanian, and Greek were most prevalent, i.e., Tosk, although the Macedonian may have affected Albanian in some border regions. The preservation of (i) vet as a possessive in Geg corresponds to the preservation of svoj – albeit with reduced functionality – in adjacent or co-territorial Slavic dialects.Footnote 73 Romani innovations have been, for the most part, system-internal, although local dialects do show contact-induced change.
6.1.1.2.4.4 Doubled Possession
There is a detail in the marking of possession on which several of the languages agree. That is, in Macedonian and Romanian, and possibly Greek as well, doubled structures occur in which possession is redundantly signaled twice, throwing emphasis on some part of the possessive structure. For Greek, there is a lone example in Thumb 1910: §143.1, taken from one of his texts (Ia24, l. 23), unfortunately of unmarked dialect provenance. The example is εμάς η αγάπη μας ‘of-us the love our,’ i.e., ‘our love,’ with emphasis thrown onto ‘our,’ and it involves the strong form of the genitive pronoun, εμάς, co-occurring with the usual weak genitive possessive μας. Thumb suggests that the emphasis-marking function of (a sort of) object reduplication (see §7.5.1) may be at work here, but with object reduplication the usual assumption is that the strong pronoun represents the canonical (underlying) structure and the weak object pronoun is the doubler, whereas here the weak possessive is the more likely candidate for being the canonical structure and the strong pronoun is doing the doubling. In any case, such doubling is not possible today, though Thumb may be giving a glimpse into nineteenth-century colloquial or regional Greek.
In Macedonian and Romanian, the doubling structure is more usual, so that what they show may be completely unconnected to whatever the Greek example may represent. Colloquial Macedonian allows the co-occurrence, along with weak dative pronoun possessives – thus just with kinship terms, as noted in §6.1.1.2.4.2 – of the possessive adjective; thus alongside tatko mi ‘my father,’ tatko mi moj (lit., ‘father to-me my’), is possible, putting more focus, interestingly, on the possessed (here, ‘father’) even though it is the possessor that is doubled. A slightly different sort of doubling occurs in Bulgarian; (6.28) is from a blog reporting on an interview with the popular singer Beloslava, so it is from a conversational register:
(6.28)
Imam svobodata v mojata si strana da pravja kakvoto iskam have.1sg freedom.def in my refl.dat country dms do.1sg whatever want.1sg ‘I have the freedom in my country to do whatever I want.’
This doubling is interesting first of all because the dative and the possessive do not formally agree. Second, the use of si adds focus on the emotional attachment the speaker has to her country; in that way, it is like the Macedonian tatko mi moj example with its focus on the possessed, not the possessor.Footnote 74
Romanian has doubling with some of its possessive structures to a greater degree than in Macedonian. For instance, the agreeing possessive affix may, according to Nicolae 2013: 342, “be doubled by full (nominal or pronominal) possessive D[eterminer]P[hrase]s in the genitive case,” and he gives the following examples:
a.
măsă-sai lui Ioni mother-his poss Ion ‘Ion’s mother’ (lit., ‘mother-his of Ion’) b.
tac-sui eii father-his his ‘his father’ (lit., ‘father-his of him’)
Moreover, in Early Modern Romanian, one other possessive strategy, the adnominal weak object pronoun type, could also occur in a doubling structure, as in the following, from works printed by Coresi, from the sixteenth century (Nicolae 2013: 346):
(6.30)
înfrumuseţarea sufletului=şi săui adornment.def soul.def.gen=refl.3sg his/her.m.sg ‘the adornment of his/her own soul’
Given the somewhat sporadic nature of these doubling structures, at least as far as Greek and Romanian are concerned, it is not clear if any real Balkanological point can be made; but as with other such parallels, the fact of superficial similarity across languages can contribute to a sense of convergence whatever the source may be.
6.1.1.2.5 Ethical Dative
Although the primary function of the dative case has here been taken to be the marking of indirect objects and the marking of possession a secondary, but still important one, these two uses do not exhaust the ways in which the dative case is of interest to Balkan linguistics. In fact, one further function of datives proves to have considerable Balkanological import, namely the so-called ethical dative (ED).
The ethical dative is a dative used to show the interest or concern that a participant in the speech act, often the speaker him-/herself, has in the matter that is at hand in the discourse. It can be a weak interest or a strong interest and may convey a certain degree of intimacy, but it reflects an emotional interest or involvement that reflects more than just the simple expression of the matter at hand. For instance, a common greeting in Greek is τι κάνεις? ‘How are you?’ (lit., ‘what are.you.doing’), but with a weak genitive pronoun referring to the speaker, τι μου κάνεις, literally ‘what for/on.me are.you.doing,’ a higher degree of concern on the part of the speaker or of presumed intimacy with the interlocutor is being expressed.Footnote 75 The same usage is seen in Macedonian kako si mi? ‘How are you?’ (lit., ‘how you.are to.me’).
Ethical datives occur in all the Indo-European branches represented in the Balkans, including Indic, via Romani. However, Akanova 2013: 100, Footnote Footnote 17 reports that Romani ethical datives “appear to be calques on the contact languages, e.g., Skopje Arli džal peske corresponds exactly to Macedonian odi si ‘go one’s way.’”Footnote 76 These datives occur in related languages outside of the Balkans, e.g., German, through the use of the dative pronoun mir ‘to/for me,’ for instance, or Russian, through the use of mne ‘to/for me’ (see Akanova 2013 for discussion and examples). And, they occur in the earlier stages of several of the Balkan languages, in particular, in Ancient Greek (Smyth 1920: §1486), in Latin (Hale & Buck 1966: §372), and in Old Church Slavonic (Večerka 1989: 263). In all of these languages, ethical datives are restricted to pronouns, and for languages with a strong-vs.-weak dimension to their pronouns, ethical datives are found only with the weak object pronouns (like Grk μου and Mac mi in the examples above).
Thus the presence of ethical datives in modern Balkan languages would seem to be best treated as an inheritance from their earlier stages, and more generally a part of their Indo-European heritage. Nonetheless, their use in Balkan Slavic is considerably more robust than in North Slavic (Akanova 2013: 109–136), a fact that is suggestive of at least some Balkan commonality to the use of the ethical dative that goes beyond what might be expected in a case of common inheritance.
Of particular interest in this regard is the use of multiple ethical datives, with both first person and second person weak object pronouns (with reflexive possible in Macedonian) cooccurring in the same sentence. This construction is found in Albanian, Balkan Romance, and Balkan Slavic, sensu stricto, but interestingly not in BCMS, even though the single ethical dative construction does occur in BCMS.
An example from Albanian is the following (from Newmark et al. 1982: 26–27):
(6.31)
Pa më t’ u e bë ai fiku well me.dat you.dat nonact it.acc make.3sg.aor that fig.def sa një shtëpi e madhe e pa më t’ i bënte as a house pc big and well me.dat you.dat them.acc make.3sg.impf ato kokrrat na sa një ftua these seed.pl exclm as a quince ‘That fig tree just grew (“for me for you”) as big as a house, and produced (“for me for you them”) figs, wow … the size of a quince.’
The feature to note is the co-occurrence of the weak object pronouns më and të; the intensified involvement reflects the speaker’s excitement at the way the figs grew. Examples include (6.32) from Bulgarian and (6.33ab) from Macedonian, drawn from examples reported on in Palmer 1997 and Akanova 2013: 96:
(6.32)
a.
Pisne onaa mi ti cigulka, šte zapišti onoj screech that me.dat you.dat violin fut squeal that mi ti klarnet! me.dat you.dat clarinet ‘That violin would screech (on you on me), and the clarinet started to squeak (on you on me)!’ (cf. Konstantinov 1967: 70) b.
Da mi si gi kupiš knigite što ti trebaat dms me.dat self.dat them.acc buy.2sg books that you.dat need.3pl ‘Buy the books that you need (on me, on yourself).’
The ‘on you on me’ and ‘on me, on yourself’ in (6.33ab) reflect the difficulty of rendering into another language the effect of this multiple dative construction. Example (6.33b) is noteworthy too for showing that reflexive dative pronouns are possible in Macedonian in this construction (although some speakers find it odd); the reference is still (vaguely) to an interlocutor, here second person, but nonetheless there is use of the reflexive, instead of the overt second person form..
Examples from Balkan Romance are given in (6.34), from Romanian, with (6.34a) drawn from Akanova 2013, and (6.34b) from Mihail 2013: 152 (with further examples in Vasilescu 2013: 397):
a.
Mi ƫi-l înşfacă cu dinƫii de cap. me.dat you.dat-it.acc grabbed with teeth of head ‘She grabbed his head with the teeth (on me on you).’ (Ion Creangă Harap Alb) b.
Mi ƫi-l ia şi-l înghite. me.dat you.dat-it.acc takes and-it.acc swallows ‘He lifts him and swallows him’
The distributional facts of the multiple ED construction are telling in shedding light on its origins within the Balkans. In particular, it is found in non-Balkan Romance, being reported, for instance, for French by Jouitteau & Rezac 2007, and it does not occur in non-Balkan Slavic (e.g., not in Russian, as Akanova 2013 reports). Based on these facts, Akanova 2013: 106 quite plausibly hypothesizes that “that the construction might have arisen in Vulgar Latin and entered Albanian and Balkan Slavic from Balkan Romance.”
It should be clear from the examples and the discussion that the ethical dative is strongly tied to speaker-to-speaker interaction. Many treatments of ethical datives emphasize their colloquial nature (e.g., Ernout & Thomas 1951: 72 on Latin, Akanova 2013: 145 on the Balkans). Moreover, they serve a clear expressive function. Akanova 2013: 144 explains this function in this way:
The expressive nature of EDs allows them to combine readily with other linguistic devices that increase expressiveness of a sentence or utterance. For instance, Macedonian ED sentences with admiratives [see §6.2.5] are perceived as more felicitous than those without them. Using present tense forms for increased narrative vividness in the past has a similar effect.
As a result of this expressiveness, the ethical dative in the Balkans, and especially the highly intimate double dative construction, can be seen as a constructional reflex of the same forces that shaped and led to the ERIC loans that are discussed in detail in §4.3 for the lexical level. Just as ERIC loans are conversationally based, so too can the ethical dative be treated as having such a basis. That so many of the Balkan languages would show increased use of an inherited construction and would adopt an innovative multiple realization of ethical datives is not surprising, given the pervasiveness of ERIC loans in the Balkans and of the forces that promoted their emergence and their spread.
6.1.1.2.6 Genitive and Dative Developments: Summation, Assessment, and Causation
A significant number of developments occurred in the Balkan sprachbund with regard to the genitive and dative cases and their typical functions. The key one, the genitive/dative merger, needs to be viewed in the various languages against the backdrop of changes in their respective case systems altogether as it fits into each language’s system in a different way (see §6.1.1). Balkan Slavic lost case altogether, so that the genitive/dative merger is part of a larger movement away from case marking per se; later Greek simplified the case system of Ancient Greek beginning in Postclassical times and has continued to weaken the role of the genitive case in and of itself, but nonetheless has retained a case system into the contemporary language; Balkan Romance still has a rudimentary case system, especially when the role of definiteness marking is taken into consideration, and alone among the Romance languages has kept certain distinctions; the same can be said for Albanian in that with definiteness figured into the mix, there is a robust case system even with a genitive/dative merger; and Romani has reconstituted the early Indic case system.
Thus, there are differences among the languages with regard to how the particular merger of genitive and dative cases has come to be and is realized in each language. Moreover, it must be noted that the case merger in itself is not surprising, inasmuch as other Indo-European languages show signs of syncretism between genitive and dative; such is the situation in Tocharian, and within the recorded history of Sanskrit there is a move towards the use of the genitive in canonically dative-like functions such as indirect object. And, the directionality of the developments like those in Sanskrit can be justified by thinking in terms of a recipient coming to be in possession of an object, with genitive as the natural case for possession. Moreover, from a Proto-Indo-European starting point with eight cases, other case mergers occurred throughout Indo-European, such as dative and locative in Greek, or ablative, locative, and dative in Germanic.Footnote 77 From all these considerations, one could reasonably surmise that the merger itself may not be due to language contact. Nonetheless, some individual aspects of this overall genitive-dative merger as well as specifics in its realization in the various languages do appear to be contact-related.
Asenova 2002: 84, for instance, makes the point that although the merger of genitive and dative cases has language-internal motivations and explanations in each of the relevant languages, these tendencies could only have been strengthened by multilingual language contact. To this we can add that details such as the loss of synthetic dative forms and their replacement with preposition + accusative in northern Greek and eastern Macedonian (as well as Bulgarian) and the shift of dative marking from the article to the preposition a[l(i)] in Ohrid-Struga Aromanian point to a continuation of the tendency toward case simplification during later contact periods.
Moreover, it is striking that there is a matching between Albanian and Balkan Romance in the differential directionality of the merger between dative and genitive, with one sort of merger in the singular versus a different direction in the plural. Given everything else that is known concerning parallels between just these two languages, on the grammatical level (see §1.2.1, §1.2.1.4, §6.1.2.2.1.2) and lexical level (see §4.2.1.1), and the possibility of a substrate being responsible for them, this merger could be added to the list of features with such a substrate basis. That is, the directionality differential is a particularity that might well point to a shared (pre-)history here at least as far as Albanian and Balkan Romance are concerned.
Thus there are aspects of the merger that have Balkanological significance, even if there are independent historical threads to reckon with as well. As for other phenomena related to the genitive and the dative, there is perhaps even more that is particularly relevant for Balkan linguistics. For instance, both Balkan Slavic and Balkan Romance have prepositional marking that moves from dative-like uses to genitive-like uses (BSl na, BRo a and la), and these developments appear to be roughly contemporaneous as far as the chronology within each tradition is concerned. More tellingly, there is clear Balkan Slavic influence at work in the genitive/accusative merger found in northern Greece for indirect object and in the use of a preposition ‘from’ (Grk από and BSl od/ot) for possession. And once one moves into further matters involving possession, as with reflexive possession, and with details of the use of dative pronominal forms in certain ethical dative constructions, the effects of contact in shaping convergent structures become even more evident.
6.1.1.3 Zero Marking
Several phenomena relevant to Balkan nominal morphosyntax come under the rubric of “zero marking,” i.e., the indication of an ostensible case function without any overt marking, essentially through the paratactic juxtaposition of one nominal as an adjunct to a controlling element, another noun or a verb, in a phrase or clause. This phenomenon is found in various of the languages and is realized in different ways, with the indication of goal of motion, location, partitivity, and instrument being especially relevant. Such zero-marked nominal usage would seem to be connected with the general Balkan reduction or outright loss of case. That is, prepositionless case-marked nominals could occur in some functions in the earlier stages of the modern languages so that loss of case marking for such nominals would result in zero-marked nominals.Footnote 78 This means that this phenomenon is of Balkanological interest even if specific points of convergence are due to independent developments within each language. Still, some uses appear to be innovations; in Old Church Slavonic, for instance, for at least some of the functions of interest here, there are no instances of case-marked nominals occurring without a preposition. In such instances, therefore, there is the possibility of discerning language contact effects, as the surveys in the following sections show. Moreover, differences in this instance between Aromanian and Romanian are instructive for suggesting a contact basis, since Romanian makes use of a preposition where Aromanian has zero marking; note especially §6.1.1.3.1 and §6.1.1.3.3.
6.1.1.3.1 Goal and Location
With certain motion verbs, a zero-marked nominal may be used as a goal in several of the languages. Thus in Macedonian, odam Bitola ‘I go to Bitola’ is possible with no overt marking for the goal (Asenova 2002: 87). Similarly, Balkan Romance shows the same pattern, as seen in examples like Aro Mi duc Bitule ‘I go to Bitola’ (lit., ‘me I.lead Bitola’). According to Koneski 1981: 128, the Aromanian is the source of the Macedonian.Footnote 79 Romanian here has a preposition la marking the goal, as in Mă duc la Bitule ‘I go to Bitola,’ which, interestingly, matches both Eastern Macedonian usage, e.g., Odam v Bitola in the use of an overt preposition, and Bulgarian usage too. Old Church Slavonic did have bare accusative nouns but only in the marking of extent of time or space or in a few specific time references, not for the goal of motion; for that function, prepositional marking was required. The Aromanian zero-marked goal-of-motion usage, on the other hand, represents the preservation of the Latin accusative of motion (Asenova 2002: 87; cf. also Rosetti 1964a: 149). These facts would indicate, as Koneski 1981: 128 has suggested, that the western Macedonian zero-marked nominal is due to Aromanian influence. Similarly, one can surmise, moreover, that the Romanian prepositional usage could be due to contact with Bulgarian. Prilep Romani usage is consistent with western Macedonian ka džas Skopje ‘we’ll go to-Skopje’ (Elšik & Matras 2006: 230). There is also the Balkan Romani tallava song Italija me ka džav ‘I will go to Italy’ (lit., ‘Italy I will go’).
This bare nominal usage extends also, in colloquial Macedonian, to the indication of location, e.g., Tetovo sum, na teren ‘I’m in Tetovo, in the field’ (Prizma 2015: 7e1; Vistinomer 2016). So also, Greek has είμαι σπίτι ‘be.prs.1sg [at] home,’ είμαι Αθήνα ‘be.prs.1sg [in] Athens,’ alongside the prepositionally marked είμαι στο σπίτι ‘be.prs.1sg at.the home,’ είμαι στην Αθήνα ‘be.prs.1sg in.the Athens,’ and in Aromanian one finds (Vrabie 2000: 67) Noi stăm Sărúnă ‘We live [in] Thessaloniki,’ as opposed to Romanian prepositional syntax here, Noi stăm la Salonic ‘We live in Thessaloniki.’ Romani can also have a zero locative, e.g., Ka muli mi daj, muli Čîrpîköz ‘When my mother died, she died [in] Çırpıköz’ (Cech & Heinschink 1999: 77, cf. also Elšik & Matras 2006: 230).
Modern Greek shows the goal usage, too, e.g., πάω Αθήνα ‘I am going (to) Athens’ (Joseph & Philippaki-Warburton 1987: 146), a form which has a distinctly colloquial flavor to it.Footnote 80 Prepositional marking, with σε, is possible here too, e.g., πάω στην Αθήνα.Footnote 81 This construction in Greek is probably an independent development, unconnected to that found in Aromanian or Macedonian, a conclusion which is all the more likely since in Ancient Greek, in poetry at least, bare accusatives could be used in the expression of the goal of a verb of motion (Smyth 1920: §1588). Thus this construction was available in Greek from earliest times, and so may be only accidentally convergent with what is seen in other languages.
There is also a tendency to eliminate case marking in locational postpositions in WRT (Friedman 2006a), as in (6.35a) compared with the StTrk (6.35b):
a.
ürti üsti kedi-ler blanket top cat-pl b.
‘On top of the blanket [there are] cats.’ örtü üstü-n-de kedi-ler blanket top-its-loc cat-pl ‘On top of the blanket [there are] cats.’
6.1.1.3.2 Partitivity
Bare nouns occurring adnominally after another noun also admit of a partitive interpretation (Sandfeld 1930: 109). In one type, the first noun is a container and the bare adnominal the material that fills it, e.g.,:
(6.36)
Bulgarian čaša voda ‘glass [of] water’ Macedonian čaša voda ‘glass [of] water’ Greek ποτήρι νερό ‘glass [of] water’ Albanian gotë ujë ‘glass [of] water’
By contrast, Balkan Romance here has prepositional marking:
(6.37)
Aromanian skafã di apã ‘glass of water’ Romanian o ceaşcă din ceai ‘a cup of tea’
In another type, the first noun is in a part-to-whole relationship with the second noun, as in:
(6.38)
Albanian një copë bukë ‘a piece [of] bread’ Greek μια φέτα τυρί ‘a slice [of] cheese’ Macedonian edno parče leb ‘a piece of bread’
For Greek and Balkan Slavic, at least, some of these uses represent innovations away from the use of the genitive in earlier stages of the language (Thumb 1910: §50c.2; Koneski 1981: 107). For Ancient Greek, the earlier usage falls under what Emde Boas et al. 2019: 30.28 refer to as the “attributive genitive,” a broad category which includes the expression of a relation of the contents to a given noun.Footnote 82 Two points are of interest here from a Balkan perspective. First, in northern Greek dialects, according to Thumb 1910: §44.2; §161.6.2, “the genitive has altogether been pushed into the background by από.” The use here of από, with the core meaning of ‘from,’ thus parallels western Macedonian uses of od ‘from.’ While the Greek developments are undoubtedly connected with the general receding of the genitive plural case in Greek, the innovative extension of από in place of genitive singular forms in the north in particular is suggestive of a contact-related development. Second, with regard to the nonprepositional type of (6.36/6.38), while it is found in Medieval Greek as early as the fourteenth/fifteenth centuries (Holton et al. 2019: 1954ff.), there is no immediate Ancient Greek source for it. Ancient Greek did have a use of the accusative traditionally referred to in the literature as the “accusative of respect” (Emde Boas et al. 2019: 30.14); however, it mainly occurred with verbs. The bare nominal specifier of content as in (6.36/6.38) thus could represent an extension into nominal syntax of this verbal accusative of respect from a structural standpoint, and indeed, the label “accusative of respect” given by Holton et al. 2019: 1954 suggests such an interpretation. However, it must be noted that the function is different; in particular, the Ancient Greek usage is tied to inalienable possession (Romagno 2017). Thus a contact effect cannot be ruled out for this Medieval and Modern Greek usage, though admittedly the directionality of influence could be from Greek onto the other languages or vice versa. The Slavic usage is attested from the twelfth to thirteenth centuries and is well established by the sixteenth (Koneski 1981: 128), so the chronology of the Balkan Slavic and Greek developments are somewhat parallel.
6.1.1.3.3 Instrument
In addition to the motion-related and partitival uses of a zero-marked nominal, instrumental uses can be found in the Balkans, mostly representing developments from earlier prepositionless case-marked nominals. For instance, in Macedonian one finds:
a.
go poseavme poleto pčenica it.acc sowed.1pl field.def wheat ‘We’ve sown the field with wheat’ b.
go služi rakija him.acc served.3sg brandy ‘S/He served him [with] brandy’Footnote 83
where the bare nominals pčenica in (6.39a) and rakija in (6.39b) correspond to uses of a prepositionless instrumental case in Old Church Slavonic (Koneski 1981: 128, 162). Similarly, in Greek one finds:
(6.40)
γέμισα το δωμάτιο λουλούδια filled.1sg the.sg room.sg flower.pl ‘I filled the room [with] flowers
where prepositional marking is an option (γέμισα το δωμάτιο με λουλούδια ‘filled.1sg the room with flowers’) just as in Macedonian the preposition so ‘with’ could occur before pčenica and rakija in (6.39). For this instrument function, which continues a use of the Ancient Greek prepositionless genitive or dative, Thumb 1910: §50c.2 cites the possibility of using a prepositional phrase for the instrument headed by από ‘from,’ e.g., γέμισα το σπίτι από γυναίκες ‘filled.1sg the house with (“from”) women.’ Romani of Agía Varvára in Greece shows a similar zero-marked construction, e.g., pherdem mi póski pares ‘I.filled my pocket money,’ i.e., ‘ … with money’; Messing 1988: 105 suggests this Romani usage is based on the Greek construction like that in (6.40), for which the alternative ordering of the noun phrases is possible.
Finally, in Aromanian too, an instrumental bare nominal is found; Vrabie 2000: 68 cites (6.41), for instance:
(6.41)
Va ti hrănéscu záhari fut you feed.1sg sugar ‘I will feed you with sugar.’
The central Balkan character of this locution is shown by the rendering Vrabie gives of this sentence into Romanian: O să te hrănesc cu zahăr ‘fut dms you feed.1sg with sugar.’ As in the instances of zero marking discussed in §6.1.1.3.1, Romanian has prepositional syntax here: the means by which the feeding takes place, ‘sugar,’ is marked by the preposition cu.
6.1.1.4 Vocative
As discussed in §4.3.5, the vocative is of considerable Balkanological interest. The focus in that section is on the vocative as a representative of a class of lexical items widely borrowed in the Balkans, and accordingly, some of the details are given about the origins of particular vocative markers through borrowing. Still, there is more to be said about the vocative, in particular in terms of how it fits into Balkan nominal systems and how it fits in with the developments that these systems underwent on their own and under conditions of contact. More generally, just as the lexical side of the vocative is revealing about language contact in the Balkans, so too is the morphosyntactic side.
The basic observation about the vocative is that it survives in all the Indo-European Balkan languages, although it is not formally distinct in Albanian and most Modern Greek noun classes. Some argue (Qvonje 1986: 69) that this preservation is a shared archaism, reinforced by contact.Footnote 84 If the vocative is considered part of the nominal case system, its preservation would seem to run counter to the tendency toward case-loss/case-reduction and the development of analytism in Balkan nominal systems. However, the vocative is clearly a different type of inflectional category from case sensu stricto. As Bühler 1934: 28–31 (in Goodwin 1990: 34–37) first pointed out, and as elaborated by Jakobson 1981 and Stankiewicz 1986b, the vocative serves to draw the attention of the addressee to the utterance and its content. What it does not do is define any sort of relation holding among elements in an utterance, a function which is found with all the true cases. The vocative is thus extra-syntactic (cf. Masica 1991: 239), and in this sense it belongs to a functionally defined category of appellative, which includes the imperative as well as vocative and hortative particles.Footnote 85 In terms of the communicative approach taken here, in which contact in its social dimension is considered, it is reasonable that a nominal form that is tied closely to communicative and conversational interaction would be more likely to survive and even spread in a context of close, intimate, and sustained verbal interaction among speakers of different languages.
The specifics of that survival in the various languages are detailed in the following sections, with an indication of aspects of the vocative that are of Balkanological interest.
6.1.1.4.1 Balkan Slavic
In Balkan Slavic, distinct vocative forms, i.e., ones that are not syncretic with other cases, occur only in the singular, with one exception: the Macedonian plural deca ‘children’ has a vocative deco ‘O children!,’ a form that does not occur in Standard Bulgarian. This general restriction to the singular reflects a feature of the vocative in Proto-Indo-European and in Common Slavic, as reflected in Old Church Slavonic, in which only in singular nouns are vocative forms to be reconstructed that are distinct within their paradigm. For substantives, the endings in Balkan Slavic are -u (limited to some masculine stems ending in a consonant), -o (mostly nouns in -a, but some in a consonant), -e (both types, often with hypocoristic value in feminines), and -Ø (nouns in -a in Turkish loans and also in the Bulgarian of Bessarbia in contact with East Slavic).Footnote 86 The zero vocative in Turkish loans in -a is essentially the Turkish stem in its Turkish form, e.g., jabandžija/jabandži! ‘foreigner nom/voc’ (cf. StTrk yabancı, WRT yabanci). It does not occur in the Torlak BCMS dialects, which in this respect go with the rest of BCMS and use -o, e.g., jabandžijo! (Radić 2001: 154), a form that can occur as a variant of jabandži in Macedonian and is recognizable to Standard Bulgarian speakers. Adjectives ending in a consonant that modify masculine vocatives can add -i. The actual rules of distribution differ among the Balkan dialects and standard languages and, for the most part, need not concern us here.Footnote 87 The vocalic endings all continue inheritances from Common Slavic.Footnote 88
Finally, an important fact here is that the vocative inflection is becoming obsolete in parts of Balkan Slavic, especially urban areas. This process was already underway in the nineteenth century (and perhaps before) as seen in folktale collections such as that of Marko Cepenkov 1972a from Macedonia. According to Koneski 1967: 240, the obsolescence of the vocative was reinforced by the fact that Turkish has no formal vocative, and Turkish loans when used vocatively keep their zero form, even when the nominative is morphologically adapted, as with jabandžija/ jabandži! above, to which can be added kardaš ‘O friend!’ and pezevenk ‘You pimp!,’ among others (cf. Turkish kardaş, pezevenk (Jašar-Nasteva 2001: 214–215)). In more recent times, the vocative has become marked in the Balkan Slavic standard languages as rude, abrupt, (excessively) intimate, and/or rural owing to its direct, appellative nature, and, as a result, its overall use is diminishing. This is especially true of the use of -o in most of Balkan Slavic, though the vocative is still felt to be normative in Standard BCMS (cf. R. Greenberg 1996b: 195).
6.1.1.4.2 Greek
In Greek, as in Slavic, the vocative is inherited from Proto-Indo-European, as far as the category and often also the form are concerned. As might be expected, there has been considerable evolution with both category and form on the way to Modern Greek. Modern Greek has a nonsyncretic vocative form only in the singular of most masculine ο-stem nouns, where the vocative ends in -ε, e.g., vocative φίλε ‘friend!,’ distinct from nominative φίλος, accusative φίλο, and genitive φίλου. In all other noun classes, a vocative form occurs that is identical to some other form in the paradigm: to the accusative in non-ο-stem singulars, e.g., γέροντα ‘old man!’ (cf. nominative γέροντας, accusative γέροντα) and even in some ο-stem nouns (disyllabic proper names, e.g., Νίκο!, a few disyllabic common nouns, e.g., γέρο! ‘old man!,’ and diminutives in -άκος, e.g., φιλαράκο! ‘little (i.e., dear) friend!,’ and to the nominative in the plural, e.g., φίλοι! ‘friends!’ / γέροντες! ‘old men!’ (versus accusative φίλους / γέροντες, genitive φίλων / γερόντων ‘friends / old men’).
This modern distribution of vocative forms shows the effects of a wide-ranging shift away from Ancient Greek. In that earlier stage of the language, there were several noun types that had a vocative form that was distinct from all other cases in the singular, in that the vocative typically had no ending, as in βασιλεῦ ‘O king!’ (versus nominative βασιλεύς), or showed a vowel difference in the last syllable, as in γέρον ‘O old man!’ (versus nominative γέρων), often accompanied by an accent retraction, as in ἄδελφε ‘O brother!’ (versus nominative ἀδελφός). In keeping with the Proto-Indo-European restriction of distinct vocative forms to the singular, all Ancient Greek duals and plurals, as well as neuters of all numbers, had a vocative identical to the nominative (as in OCS).
From a diachronic perspective, Greek, like Balkan Slavic, has shown a general retreat of the vocative as a distinct form. Even with the syncretism of the vocative with other paradigmatic forms in Ancient Greek, there were still many vocatives that were formally distinct and, importantly, there were considerably more than there are in Modern Greek. Moreover, the accent retraction associated with many vocatives has been eliminated, so that the modern vocative of ‘brother’ is αδερφέ, not the expected outcome άδερφε*.
Overall, then, the facts about the form and the accent can be taken as an indication of the diachronic weakening of the vocative as a category. Still, one expansion of the vocative can be noted, namely the innovative creation sometime in Postclassical Greek – the chronology is uncertain but the forms exist in Modern Greek – of the masculine vocatives αυτέ and τέτοιε to the third-person pronoun αυτός ‘he (she, it, they)’ and the qualitative pronoun τέτοιος ‘such,’ respectively; as described by Householder et al. 1964: 84, these forms “are sometimes used when the name of the addressee is unknown as in αὐτέ, πῶς σὲ λένε? ‘You, what’s your name?’.”
Thus the story of the vocative in Greek is like that in Balkan Slavic – retention of the category and the form, to some extent, but with some clear receding. These could well be independent developments, but the retention of the accentual distinctness of the vocative outside of the Balkans (cf. Pontic Greek άδεφλε) suggests that the parallel paths for Greek and Balkan Slavic here were not completely independent.
There is one further point to be made regarding the vocative in Greek. Various Greek dialects preserve the Ancient Greek interjection ὦ ([ô:]) that can precede vocatives; as described by Margariti-Ronga & Papadamou 2019a: 138–139, “it has survived in the dialects of Lesbos, Naxos and Siatista” and they consider the “core of its geographic distribution to be Epirus (Sarakatsans, Delvino, Ioánnina, etc.), Western [Greek] Macedonia (Germa, Kotyli, Katafygi, Siatista, Galatini, Western Voio, Avgerinos, etc.) and the western part of Central [Greek] Macedonia (Roumlouki, Mountainous Pieria, etc.).” This distribution suggests that Albanian, too, could have played a part in the preservation (see §6.1.1.4.3, and also Papadamou 2018).
6.1.1.4.3 Albanian
For the vocative, Albanian uses the indefinite nominative form of the noun, with or without a preceding vocative particle such as o, bre, more (m), moj (f).Footnote 89 The particle o, which is stressed and has a distinct rising intonation, is especially common as the unmarked indication of vocative. However, there are certain lexical items with special distinct vocative forms of their own, e.g., biro ‘O son!,’ although proper names can also occur in this form, e.g., Agimo! (Buchholz & Fiedler 1987: 215); the specific shape of this marking, in -o, has been attributed to Slavic influence (cf. §4.3.5). Albanian can also use the marker of the plural imperative (also second person plural present) in the verbal system as a vocative plural marker, e.g., shok ‘friend,’ shokë ‘friends,’ shokëni ‘O friends!.’Footnote 90 Unlike the situation in the other Balkan languages, the definite form of the noun is used if there is an attribute that requires definiteness (e.g., certain possessives), if a collective of individuals is addressed, or if a preposed (and thus expressive) adjective is used: Dëgjoni, vajzat e mia! ‘Listen, my girls!,’ O malet e Shqipërisë e ju o lisat e gjatë! ‘O mountains of Albania, and you O tall oaks!,’ I mjeri atdhe! ‘O unfortunate fatherland!’ (Buchholz & Fiedler 1987: 215). However, we should note that Albanian, and for that matter, all the Balkan languages, like some non-Balkan languages (e.g., French), can also use definite substantives as vocatives, though these usages are mostly expressive or found in folklore, e.g., Alb Çuni, eja këtu ‘Hey boy, come here’; cf. Blg Ej, krivokrija ‘Hey, crooked-legs!,’ Rmn Draga mea, mândra mea! ‘O my dear, O my beauty!,’ Grk Γεια σας, χαρά σας, μάστοροι, και σεις οι μαθητάδες ‘Greetings, O masters, and you, O apprentices’ (Asenova 2002: 135–136).
Not enough is known, or is knowable, about the prehistory of the vocative in Albanian to allow for much to be said here, but the vocatival type with a particle plus a noun is consistent with the general move towards analytism seen in the Balkan noun (see §6.1.1) and it seems clear that contact with Slavic has played a role in the development of the vocative.
6.1.1.4.4 Balkan Romance
The background for the Balkan Romance vocative is the situation in Latin, in which there was a distinct vocative form only in the singular of the masculine -u- (prehistoric *-o-) stem nouns; in the singular of all other stem-types, and in all plurals, the vocative had the same form as the nominative. In Balkan Romance, a variety of means of marking vocatives can be found, representing innovations from the Latin starting point.
In Romanian (Graur et al. 1966: 78–81; Croitor & Hill 2013), the vocative suffixes are -e and -ule for masculine singular, -o for feminine singular, and -lor for the plural of both genders. The standard language permits vocative only for persons and personified inanimates, but no such restriction exists in colloquial usage, where, e.g., cuc/cuce ‘cuckoo’ and codru/codre ‘forest’ occur. The suffix -ule is affective (diminutive, ironic), and in general the vocative denotes intimacy. These suffixes are not obligatory, in that zero-marked vocatives are possible, so that, for instance, both prieten and prietene for ‘friend!,’ and om and omule for ‘man!’ are possible. As Coteanu 1958 notes, the definite can also serve as a vocative, e.g., fiul meu! ‘O my son!,’ and he argues that this is the origin of -ule. The suffix -lor in the plural “is syncretic with Genitive-Dative Case of the definite declension” (Croitor & Hill 2013: 805), and represents a move away from the Latin vocative-nominative syncretism in the plural.
There are as well various vocative particles that, like those of Albanian (see §6.1.1.4.3), can be viewed as “lexical (pragmatic) markers of addressee” (Croitor & Hill 2013: 804). These include mă(i), bă(i), bre, and fă and in the combination of these markers with vocative nouns, the noun “may be Case marked or not” (idem), so that (marked) măi fato and (not marked) măi fată are both acceptable for ‘girl!.’Footnote 91
In both Aromanian (Capidan 1930: 383–396) and Meglenoromanian (Atanasov 1990: 195), means similar to what is seen in Romanian occur for marking vocatives. Thus -ule occurs in both, and various vocative particles are found in Aromanian, e.g., bre (as in a bre om ‘hey man!’) but also (Capidan, 394) laĭ, lḙa, ale, le. This last can surround a noun, e.g., le soră le ‘O sister!’ (ibid.) in the same way that măi can in Romanian, albeit rarely (Croitor & Hill 2013: 804), as in măi fată măi ‘girl!.’ Vrabie 2000: 48 gives vocatives in -o and -e for some nouns in Aromanian, e.g., sóro ‘O sister!’ (nominative soră), doámne ‘O sir!’ (nominative domnú).
It is generally agreed that the Balkan Romance vocative forms were influenced by Slavic (see Pătruț 1963, 1964, and sources cited in Tucker 1944); Densuşianu 1901: 1.244, for instance, refers to a Bulgarian (but actually Balkan Slavic) “suffixe hypocoristique le” (‘hypocoristic suffix le’), e.g., BSl male ‘O mama!’, goro le ‘O forest!’ (nom gora). The -o is often compared to the feminine vocative ending in OCS (and BSl), e.g., ženo (to nominative žena ‘woman’). Tucker, however, is not convinced by the claims of Balkan Slavic origin, but is able to offer only weak arguments for all but -o. As for -o, he argues (p. 25) that the Romanian “vocative ending -o does not behave at all like a part of the word” and in particular does not reverse the shift of -o- to -oa- the way the absence of -ă or -e in a following syllable generally does; cf. foaie ‘leaf’ ~ plural foi, frumoasă ‘beautiful’(f) ~ frumos (m), but vocative frumoaso (not *frumoso) ‘O beautiful (one).’ Based on that behavior, he suggests that “voc -o is in origin not a case-ending, but a distinct word used as a postpositive like the definite article” (p. 26), and points to “the familiar [Latin] interjection o, which exists in Roumanian … [and] occasional instances [of which] may be found in Latin poets” (p. 27). Tucker’s arguments are somewhat inconclusive for the Balkan Romance vocative endings other than -o, but for -o, they do raise the possibility of Romance input. However, Balkan Slavic influence cannot be ruled out even for -o. What might be closer to the truth is the suggestion of Densuşianu 1901: 1.244 of a mixed origin for this -o: “Peut-être faut-il plutôt supposer que le vocatif roumain reproduit d’un côté le vocatif latin, de l’autre côté le vocatif slave; la terminaison slave se serait superposée sur celle qu’on avait héritée du latin” (‘Perhaps one must rather suppose that the Romanian vocative reproduces on the one hand the Latin vocative and on the other the Slavic vocative; the Slavic ending would have been superimposed on that which was inherited from Latin’). That is, (-)o in Balkan Romance could be the Latin vocative particle o in its form, but always positioned as if it were identified with Slavic suffix -o. That is, it would not be an outright borrowing from Slavic nor an outright inheritance from Latin.
See §4.3.5 for a discussion of lexical vocatives in Balkan Romance, and see Schnelzer 2024 for the most recent discussion of Balkan perspectives on the development of vocatives in Romanian.
6.1.1.4.5 Romani
Unlike other inflectional markers, the Romani vocative attaches directly to the bare stem. The singular is generally unstressed -a (for nouns in a consonant) or -éja or -ája (for nouns in -o), although -o and -e sometimes also occur. The feminine is usually -[j]e or -[j]a, sometimes -o. There is a separate plural form which usually is -ále[n] or some variant thereof.
The historical developments with the vocative in Romani show a mix of retention, innovation, and contact-related effects. As to form, Sampson 1926[1968]: 131–132 compares the masculine -a and the feminine -e with Sanskrit vocatives of the same shape from -a and -ā stems, respectively. And, as to category, the creation of a productive and paradigmatically distinct plural form represents an innovative extension of the vocative and is unique in the Balkans.
What amounts to an (innovative) analytic vocative construction is the use of mi ‘my’ in vocatives such as mí phen! ‘sister!’ or mí čhaj! ‘daughter!,’ without any overt sense of possession implied through the use of the possessive. The mi is the stressed element in the phrase, suggesting a retraction of accent that is reminiscent of what occurs in Sanskrit with the vocative (cf. pitā́ ‘father.nom’ ~ pítar ‘father.voc’). It is also worth noting that in Romani, as in Sanskrit, the monovocalic vocative ending is unstressed, e.g., phral – phrála ‘brother,’ phen – phéne ‘sister,’ phurí – phuríje ‘old woman,’ and in some nouns, as a result, there is an accent retraction, e.g., devél – dévla ‘God,’ again like early Indic vocative accentuation.
In other instances, contact with other languages has clearly been decisive. For instance, in the S Vlax dialect of Agía Varvára (Igla 1996: 33), borrowings in a vowel do not distinguish vocative from nominative, while those with facultative -s in the nominative never have that -s in the vocative, e.g., nom čobáno[s], acc čobános, voc čobáno ‘shepherd.’ In Bugurdži (Boretzky 1993: 33) and Sepečides (Cech & Heinschink 1999: 19), -s is regular in the nominative and accusative and dropped in the vocative. The nominative -s itself is arguably a result of Greek influence, as is its absence in the vocative.Footnote 92 Further, in Sepeči Romani (Cech & Heinschink 1999: 22), borrowed feminine nouns in -ija have a vocative in -ijo, e.g., džadija – džadijo ‘witch’ (cf. WRT cadi, StTrk cadı), which apparently entered via a Slavic intermediary together with the Slavic vocative marking.
There is considerable controversy over the history of the innovative plural marker. The category itself can be seen as an internal development within Romani, motivated by the presence of distinct vocatives in the singular, but the precise source of the form is disputed. Sampson 1926[1968]: 131–132 treats the le[n] of the vocative plural as the Slavic vocative particle le attached to the vocative. Boretzky 1994: 93 proposes the origin of all the vocative endings in cliticized interjections (a, ja, le), noting, however, that they must be very old since they are Common Romani.Footnote 93 Matras 2002: 80 suggests an origin in old deictic or emphatic particles for forms with -l- and for -[j]a (for which he suggests the Proto-Romani particle *-a found with oblique pronouns and demonstratives), respectively.
6.1.1.4.6 Turkish
Old Turkic has a postposed vocative particle -(y)a or -(y)A but also used the nominative in vocative functions (Erdal 2004: 351, 361). In Balkan Turkish, the vocative is expressed by the usual vocative particles (see §4.3.5), and in some instances those particles have multiple potential sources that include Turkish (e.g., de, a, etc.). WRT also has the vocative suffix -o, e.g., babo ‘O father,’ abo ‘O older brother,’ etc. (Bayram 1985), arguably of contact origin, whether directly from Balkan Slavic or through an Albanian intermediary. Turkish has a diminutive suffix -o, e.g., Neco < Necat, which could have contributed to the usage, but the perceived prevalence and vocative usage of -o in WRT is arguably a Balkan feature.
6.1.1.4.7 Judezmo
Judezmo, with its lack of cases and no apparent vocative inherited from its earlier Spanish source, offers little insight into the Balkan vocative. Sources such as Varol Bornes 2008 show a considerable amount of lexical material that is vocatival in nature, involving various terms of address, but nothing grammatical per se. The relevant lexical material is surveyed in §4.3.5.
6.1.1.4.8 Conclusions Regarding the Vocative
Based on the preceding, it is clear that the vocative is retained in the various languages even if, for the most part, it is reduced as a formal category in the grammar. From a structural point of view, the very survival of vocatives is not consistent with the reduction or outright loss of other cases and the emergence of analytic structures. However, the survival of precisely this case in Balkan Slavic can be motivated by reference to the fact that it is the only nominal case directly involved in speaker-to-speaker interaction, and the emergence of vocative particles in all the languages is consistent with analytism.
Moreover, even with some retreat for the vocative, there are occasional signs of life that counter it. Examples are the innovative extension of the vocative into pronouns seen in the Greek αυτέ and τέτοιε (see §6.1.1.4.2), the creation of new plural vocatives in Romanian with the areally and familially unusual syncretism with the dative/genitive case (see §6.1.1.4.4), and a completely new plural form in Romani (see §6.1.1.4.5). The lexically particular Macedonian innovation with deco appears to show that such an extension was in principle a general possibility, but in fact this form is arguably a lexical loan, like the Church Slavonic vocative Bože ‘O God!’ in Russian.Footnote 94
Interestingly, this innovative foray into the plural on the part of the vocative is most fully developed only in the most sociolinguistically isolated speech community, namely that of Romani. And in fact, some aspects of the retention of vocatives also make sense sociolinguistically. In particular, the archaism within Romani of the maintenance of accent retraction, as in some Pontic Greek forms (see §6.1.1.4.2), is interesting since Romani speakers, of all the Balkan speech communities, have been the most socially isolated and among the least integrated into the mainstream of Balkan society, just as Pontic Greek speakers of Thumb’s late nineteenth-century account were isolated from Balkan Greek.Footnote 95
The vocative thus shows how both structural and social factors, as well as outright contact, play a role in the Balkan nominal system. Moreover, as discussed in §4.3.5, as conversational elements par excellence, they are consistent with the typology of borrowing in the Balkans, being prime instances of ERIC loans (see §§4.1, 4.2, 4.3).
6.1.1.5 Other Case Developments: Further Mergers
The various changes in Balkan case systems discussed in the preceding sections do not exhaust the range of interesting developments, as there are yet others to report on. They are of a somewhat more restricted nature, with particularly interesting manifestations in Romani and Balkan Turkish, though with some realization elsewhere as well.
Romani and West Rumelian Turkish are especially noteworthy in this regard as they are unique among the Balkan languages in having mostly preserved a rich system of cases. In the most conservative dialects of Romani, the eight-case system inherited from Old Indic (nominative, accusative, vocative, dative, locative, ablative, instrumental, and genitive) is recapitulated, albeit, as already noted, with new material.Footnote 96 Moreover, Balkan Turkic retains the inflectional cases with which it arrived in Asia Minor, from earlier stages, so that the languages at the top and the bottom of the Balkan social scale, each for its own reasons, resisted the general Balkan trend towards analyticity in nominal inflection. Still, each shows some tendencies toward simplification that reflect an intersection between the areal and typological. From a typological viewpoint, peripheral cases might be expected to be lost first, and as detailed in the following sections (and also mentioned in §6.1.1.2.1.5), it is precisely those cases that are affected in these two languages. Romani dialects of the Balkans, for instance, have developed some analytical alternatives which, as Romani increases in status owing to the relatively greater social integration of its speakers, are increasingly replacing some of the old inflectional cases, e.g., the ablative (VAF field notes for North Macedonia, Igla 1997 for Bulgaria).
6.1.1.5.1 Dative/Locative Merger
Sandfeld 1930: 191–192 remarks on Balkan languages in which two different senses of ‘where’ – the locatival “ubi,” i.e., ‘place in which’ and the goal-oriented “quō,” i.e., ‘place to which’ senses – converge on the same form. In Macedonian, for instance, kade (colloquially kaj) is used for both functions:
a. kade živeeš
where live.2sg
‘Where do you live?’ (i.e., ‘in which place … ?’)
b. kade odiš
where go.2sg
‘Where are you going?’ (i.e., ‘to which place … ?’)
This same situation obtains in the other Balkan languages, e.g., Bulgarian kăde (also gde), Greek πού, Romanian unde, Aromanian iu, Albanian ku, and Romani kaj, and extends to other adverbials, e.g., for ‘there’ (Greek εκεί, Albanian atje, Macedonian tamu, Aro aclo, etc.) and for ‘outside’ (Albanian jashtë, Bulgarian văn, Macedonian nadvor, Romanian afară, etc.). However, this merger of senses is also widespread outside of the Balkans (cf. English where (as in the glosses in (6.42)) or French où), so that even Sandfeld expresses some doubts as to whether this counts among “phénomènes essentiellement balkaniques” (‘essentially Balkan phenomena’). However, Balkan Slavic deviates from the rest of Slavic in showing this semantic merger; Russian, for instance, reflects the inherited Common Slavic situation in having gde for locatival ‘where’ versus kuda for directional ‘whither,’ and BCMS has gd[j]e (with variants) versus kamo (replacing kuda) for the same distinction (BSl kamo has a different, nondirectional and expressive meaning of the type ‘if only, let alone, etc.’). The difference between Balkan and non-Balkan Slavic is the sort of evidence that is at least suggestive of a contact-related explanation for the Balkan Slavic facts, thus making this feature relevant for Balkan linguistics, despite Sandfeld’s misgivings.
Moreover, this merger has a realization outside of the system of adverbials in various Balkan languages. For instance, as noted in §6.1.1.1.3.2, in Greek the preposition σ(ε), which governs the accusative case, is used for traditional dative functions, such as the marking of the indirect object but also place-to-which, as in (6.43ab), and as well for the expression of location, as in (6.43c):
a.
το έδωσα στον Γιάννη it.acc give.pst.1sg to.the.acc.sg John.acc ‘I gave it to John’ b.
πάμε στην Αθήνα go.prs.1pl to.the.acc.sg Athens.acc.sg ‘I am going to Athens’ c.
μένω στην Αθήνα live.prs.1sg in.the.acc.sg Athens.acc.sg ‘I live in Athens.’
And in Balkan Slavic, the preposition na marks indirect objects, location, and motion (na fakultet ‘to/at/in the faculty/department/[academic] division’) and in Albanian and Balkan Romance, the prepositions në and la, respectively, have similar dual functions for place-in-which (locative) and place-to-which (e.g., në Tiranë ‘in Tirana; to Tirana,’ la Bucureşti ‘in Bucharest, to Bucharest’). Further, in both Balkan Turkish and Romani, there is sometimes a merger of dative with locative within the nominal system, i.e., affecting the cases. The Balkan Turkish locative thus can sometimes occur with a verb of motion to indicate movement towards (cf. Friedman 2006a):
(6.44)
gittı-k Selanik-te went-1pl.aor Thessaloniki-loc ‘We went to Thessaloniki.’
In Standard Turkish, by contrast, the dative Selâniğ-e would be found here. Kakuk 1972 reports that more rarely a dative occurs where a locative is expected. Relevant here too is the fact that bare nouns, without any prepositional marking, can be used – as discussed in §6.1.1.3.1 – for both goal and location in some of the languages, e.g., Greek and Macedonian.
As for Romani, some dialects in contact with Balkan Slavic tend to merge the locative (-te) with the dative (-ke), usually in favor of the dative. Furthermore, both the dative and the locative, as well as the ablative case constructions, have come to be replaced with prepositional markers that derive from the corresponding case forms of demonstrative pronouns. Thus while some dialects make the distinction in (6.45ab), those affected by Balkan Slavic show the polysemous expression of (6.45c), where the preposition k-e (or in some dialects, t-e) derives ultimately from the postpositional element (case-suffix) -ke reinterpreted as a preposition:
a.
jekh-e aindž-a-te one-obl field-obl-loc ‘in a field’ b.
jekh-e aindž-a-ke one-obl field-obl-dat ‘to a field’ c. k-i jekh aindž
to-f one field
‘in/to a field’
While Naylor 1981 sounds some cautionary notes concerning a contact-based account of these facts, based on the fact that dative/locative mergers occur in some non-Balkan Slavic languages, e.g., Polish, the geographical distribution in the Balkans of such mergers, in the broadest sense, is telling, and highly suggestive of language contact playing some role.
6.1.1.5.2 Genitive/Ablative Merger
A further case-related development in which language contact plays a role in the Balkans is a merger of genitive and ablative in some dialects of Romani.Footnote 97 In particular, in Romani varieties in North Macedonia, a nominal construction occurs with the etymologically ablatival tar- (as a preposition) and showing polysemy between ‘of’ and ‘from’ semantics, as in (6.46);Footnote 98 many or most other dialects differentiate these two meanings formally with different case marking, as in (6.47):
(6.46) aindž-a-tar (alternatively: tar-i aindž)
field-obl-abl from-f field
‘from a field’ / ‘of a field’
a. aindž-a-tar (alternatively: tar-i aindž)
field-obl-abl from-f field
‘from a field’
b. aindž-a-ko[ro]
field-obl-gen
‘of a field’
This development in Romani shows influence from Macedonian, since the preposition od, etymologically ‘from,’ is used to mark both ablative and possessive relations (and other “genitive” uses too) in western dialects;Footnote 99 thus the phrase in (6.48) is polysemous, as indicated in the gloss:
(6.48)
knigata od Blažeta book.def from Blaže.obl ‘the book that belongs to Blaže /the book that comes from Blaže /the book that is by Blaže’
and similar genitive-like polysemy is found with the innovative ablative in Romani. This process appears to have progressed to different degrees in various Romani dialects but it is especially noticeable in North Macedonia and Bulgaria; in Skopje, modern-day contact includes significant calquing and translation phenomena in various media of a type for which Romani was not used a generation or two ago, and for the youngest generation, the ablative is already becoming archaic (Sejdo Jašarov, p.c. 2008).
6.1.1.6 Summation on Case Overall and a Consideration of Causation
The preceding sections show that as far as nominal case morphosyntax is concerned, as for nominal morphosyntax more generally (see §6.1.6), an overall move in the direction of analytism is the most significant development, even if there is much more than case to consider. Still, a key question that remains is what the causes of the emergence of analytic structures are.
Asenova 2002: 89–90 sums up arguments for the causality of (nominal) analytism, for which the logical foci are Balkan Romance and Balkan Slavic since, for one thing, Balkan Romance is the most conservative of all Romance while Balkan Slavic is most innovating of all Slavic in this respect (if one excludes the innovative synthetic marking of definiteness).Footnote 100 One hypothesis is that Balkan Slavic influenced Balkan Romance conservatism, while Balkan Romance analytism influenced Balkan Slavic, which then carried the tendency further than in Balkan Romance itself (Asenova 2002: 89–90, who cites Graur 1955: 226; Seidel 1958: 15–16; Rosetti 1964b: 52, 1968: 124). Asenova also adduces the example of Aromanian, in which the numerals 20–29 extend the apparent Slavic teens pattern of the type X-on-10 to X-on-20, e.g., unãsprãyingjits ‘21.’Footnote 101
Nonetheless, the processes of case reduction and/or case analytism are on-going, as seen, for instance, with the genitive plural in Greek (§6.1.1.1.3.3). While the two tendencies are contradictory, in that the former breaks down case distinctions while the latter functionally maintains them, they are not mutually exclusive. Although Balkan Romance in general is more conservative in its case marking, the Ohrid region Frasheriote dialect of Aromanian and also Meglenoromanian are closer to most of Macedonian in their reduction of synthetic case marking (Markovikj 2007:51; Atanasov 1990: 195–198). At the same time, however, the western Macedonian dialects are more conservative in their preservation of oblique forms for animate nouns, and the most peripheral of the southwestern Macedonian dialects (Boboshtica (Mac Boboščica) and Drenova (Mac Drenovjäne)) preserved accusative and dative case marking in the postposed definite article (and some indefinite nouns), undoubtedly owing to the influence of the definite accusative and dative postposed articles in Albanian and Aromanian (Mazon 1936: 53–64, but Steinke & Ylli 2007: 314–317), both of which were significant contact languages in the region. Peripheral Bulgarian dialects in the Rhodopes, Thrace, and northeastern Moesia are likewise more conservative in preserving remnants of case distinctions (Mladenova 2007: 115, 345), in these instances in contact with Greek, Turkish, and Romanian. On the other hand, the nominative/oblique distinction in the postposed definite article of Modern Standard Bulgarian is the invention of codifiers in the nineteenth/early twentieth centuries (Friedman 2002b; Fielder 2018). At the northwestern periphery of Balkan Slavic, the Torlak dialects of the former Serbo-Croatian show a case inventory that in the context of Balkan Slavic counts as conservative but in the context of the former Serbo-Croatian counts as the most reduced (Friedman 1977a). For Slavic, then, the picture is one of a central area of innovation from a heartland from eastern North Macedonia to central Bulgaria (see §6.1.1) with case simplification to a maximum of three cases vis-à-vis the Common Slavic heritage being nearly coterminous with the spread of the definite article (on which see §6.1.2.2.1.1).Footnote 102
6.1.2 Referentiality
The rubric of referentiality comprises topics involving different degrees to which entities are picked out and identified in discourse. Of relevance here are deixis, definiteness and the nexus of deixis with definiteness, and the polar opposite of definiteness, namely indefiniteness. Each of these topics has elements of considerable Balkanological interest. Definiteness has attracted much attention in the Balkanological literature, but the marking of indefiniteness has not figured in traditional accounts of Balkanisms, even though the chronotopic facts about each in the Balkans suggest that the two poles of definiteness are related.
6.1.2.1 Deixis: Two versus Three Distinctions
Unlike referential areal features such as object reduplication, definite/indefinite articles, and double determination, deictic oppositions as such do not generally figure in Balkan linguistic accounts, and with reason: deictic systems in the Balkans generally show resistance to contact-induced changes at the macro-level, although local level deictic Balkanisms do occur. Balkan deictic systems can be characterized as two-term (which are all simplifications from earlier three-term systems) and three-term (which continue inherited distinctions albeit with different material).
We begin with a very brief overview of the history of Balkan demonstrative systems. Common Slavic, Greek, and Latin all inherited a three-way demonstrative opposition corresponding roughly to the three persons, OCS sĭ, tŭ, onŭ, Latin hic, iste, ille, Ancient Greek ὅδε, οὗτος, ἐκεῖνος, all roughly ‘this here’ (like 1st person), ‘this/that’ (like 2nd person), and ‘that there’ (like 3rd person). Similarly, early Indic, representing the precursor to Romani, showed at least a three-way contrast (cf. Skt sa ‘that, this,’ eṣa ‘this here,’ ayam ‘this/that.’Footnote 103 The Albanoid system is uncertain, but a three-way contrast seems likely for Proto-Albanian. We then examine Turkish and Romani.
The non-Slavic Balkan languages are all consistent in their use of two-term or three-term systems. Albanian and Balkan Romance (like all of Romance, including Judezmo) have two-term proximal/neutral systems, e.g., Alb ky/ai, Aro atsest/atsel ‘this/that.’Footnote 104 Modern Greek has preserved the earlier three-way distinction of ‘this, this/that, yon,’ via τούτος (remade from the neuter of οὗτος), αυτός (from an Ancient Greek intensive reflexive form meaning ‘self; same’ – see below), and εκείνος, as has Turkish with bu, şu, and o, likewise ‘this, this/that, that.’ Romani shows an interesting elaboration on the earlier system (see below). Some of these languages deserve a deeper treatment for the full import of the deictic systems to become clear.
Thus, we start with Greek, where there is abundant attestation. While the contemporary language has a three-way contrast, as noted, early descriptions (sixteenth to seventeenth centuries) of Modern Greek, discussed below, treat τούτος versus εκείνος as a basic two-way opposition; Ancient Greek ὅδε had fallen from use by the early Middle Ages (Browning 1983: 66). The developments with αυτός are particularly important here. In ancient times, forms of αὐτός ‘self’ could be used as a simple anaphoric pronoun, only in oblique cases in Classical Greek but in the Hellenistic koine in the nominative as well.Footnote 105 This change must have involved passage through a neutral demonstrative, in the way that Clackson & Horrocks 2007: 215 describe for the Latin intensive ipse ‘self’; that is, in its development into a nonemphatic anaphoric pronoun ‘he’ in texts such as the de Bello Hispaniensi of the first century BCE,Footnote 106 it represents “a probable weakening and extension of the emphatic use where the expected demonstrative may be omitted, = ‘that very one just mentioned.’”Footnote 107 And indeed, that neutral demonstrative use emerges, or re-emerges if it continues that intermediate deictic function, in the Hellenistic usage of αὐτός as equivalent to the ancient demonstrative οὗτος, noted by Jannaris 1897: §1418b and seen in examples such as (6.49) from Egypt in 169CE (Horrocks 2010: 149):
(6.49)
ὑπέρ αὐτοῦ τοῦ πράγματος concerning this.gen.sg the.gen.sg matter.gen.sg ‘concerning this matter’
And, of course, this deictic function is seen in the modern use of αυτός as the proximal demonstrative ‘this,’ a use which co-exists with the use of αυτός, in all case forms, as the strong third-person pronoun as well, also continuing a Hellenistic function.
The Modern Greek opposition is seen also in corresponding locational adverbs εδώ/αυτού/εκεί ‘here/here ~ there/there.’ According to Mackridge 1985: 226, however, in actual usage Modern Greek is more of a two-term system:
[τούτος is] largely absent from more elevated styles (unless, paradoxically, the writer/speaker is imitating Katharevousa ούτος), it being considered to sound rather vulgar, at least when it is used about a person; the same is true of αυτού, which traditionally corresponded not to τούτος but to αυτός. A two-term system now prevails, αυτός being the unmarked form, indicating referents close to either the speaker or hearer, while εκείνος is the marked form, for referents distant from both.
αυτός also means ‘the latter’ and εκείνος ‘the former’ in discourse, or it can signal a change in grammatical subject (Mackridge, idem) when the new subject comes after the other one. Although Mackridge cites Tzartzanos 1946: 138–139, as representing an older three-term system, more recently both Householder et al. 1964: 89–90 and Joseph & Philippaki-Warburton 1987: 164 treat the three-way pronominal system as normal. Whereas Householder et al. 1964: 141–142 also treat the three-way locative adverbial opposition without comment, Joseph & Philippaki-Warburton 1987: 164 implicitly treat the demonstrative locative adverbial opposition as binary (see Table 6.11).
| τούτο εδώ/*εκεί | αυτό εδώ/εκεί | εκείνο εκεί/*εδώ |
| this here/*there | this ~ that here/there | that there/here |
It is interesting to note that in Greek, it is the neutral pronoun that functions as the filler form in expressions that correspond to English thingamajig and whatchamacallit, e.g., πώς το λένε ‘what’s it called,’ η αυτή ‘what’s ‘er name,’ πώς σε λένε ‘what’s yer name’ (Mackridge 1985: 227).Footnote 108
The Ancient Greek three-way distinction was altered during the Koine, in that ὅδε became rare and there were the aforementioned shifts with αὐτός (Browning 1983: 66). The change of reflexives to demonstratives occurs in other (non-Balkan) languages, although we cannot rule out the influence of nonnative speakers of Greek in this change. Nonetheless, given the chronology, there is nothing we can say to identify it as specifically Balkan. During the early medieval period (600–1100) , neuter τούτο- extended to the masculine and feminine, replacing οὑτο- in form. During this period, ἐκεῖνος remains but decreases in frequency. According to Horrocks 2010: 149, the modern construction with αυτός, e.g., αυτό το βιβλίο, literally ‘this (the) book’ (cf. the Albanian equivalent ky libri), has sporadic attestation in Roman-era materials such as (6.49) above. Aside from this, however, Horrocks is concerned with form, not function. At issue, then, is not the earliest attestations of the rise of αὐτός and the demise of ὅδε, which clearly pre-date the arrival of the Slavs and the Ottoman period, but rather the timing of the ecology of the selection of αὐτός as the unmarked demonstrative.
What is interesting from the Balkanological perspective here is that the shift of αὐτός from ‘self’ to a neutral demonstrative, although already underway much earlier, as the evidence in the preceding discussion shows, appears to stabilize precisely during the Ottoman period. Manolessou 2001, in her comprehensive treatment of the history of demonstratives in Greek, shows that τούτος was well represented as a demonstrative, much more so than αυτός, in the fifteenth through the seventeenth centuries; in fact, she determines that “the real turning point comes in the later eighteenth century: Until then, one may find texts with only οὗτος, or with αὐτός and οὗτος interchangeably, but there are no texts/authors using αὐτός exclusively, or in an overwhelming majority, similar to MG[rk] usage.”Footnote 109
This finding accords with the evidence of equivocation regarding αυτός as a demonstrative in several grammars of Modern Greek dating from early Ottoman times. For instance, in the grammar of Nikolaos Sophianos (c.1540), whose own Greek was relatively colloquial (e.g., with να-clauses instead of more archaic infinitives, contractions like ‘σ τήν for εἰς τήν ‘to the’), only two pronouns of the type he labels δεικτικαί ‘deictic’ are listed: τοῦτος and ἐκεῖνος. Although he does give αὐτός, he labels it ἐπιταγματική ‘subsidiary.’ Moreover, when he does make use of αὐτός in his actual texts, some of these occurrences are amenable to an interpretation of “X itself” or “the very X,” i.e., not strictly speaking a demonstrative:
(6.50)
ἡ δευτέρα θέλει καὶ αὐτὴ στον ἀόριστον τὸ η ἢ τὸ α … the.f second.f needs and autos.f in.the.n aorist.n the.n ē or the.n a ‘The second [conjugation] itself takes in the aorist [stem] ē or a … ’
His usage primarily has τούτο- with occasional instances of ἐκεῖνος. Thus, while the usage of αὐτός might be at odds with his description of the pronominal system as purely binary, it is also clear that by the first half of the sixteenth century αὐτός had not yet settled into its current state.
A couple of generations later, Girolamo Germano 1622, in his grammar of Greek written in Italian, gives three pronouns under Pronomi demonstrativi: τοῦτος, αὐτός, and ἐκεῖνος, but he glosses them, respectively, as ‘questo,’ ‘cotesto ò questo,’ and ‘quello.’ The glossing of αὐτός as ‘cotesto ò questo’ is peculiar, since, literally, it means ‘that or this,’ but the word used for ‘that’ is not the ordinary questo but rather a word considered “uncommon,” at least in modern Italian. Nearly contemporary with Germano is Simon Portius 1638, who in his chapter 5 gives only two deictics: τοῦτος and ἐκεῖνος (“demonstrativa sunt duo, τοῦτος vel ἐτοῦτος hic, ἐκεῖνος vel κεῖνος ille” (‘the demonstratives are two, τοῦτος or ἐτοῦτος this, ἐκεῖνος or κεῖνος that’)).
Thus, Greek has undergone a development in which the older three-way ‘this(specific)-this(general)-that’ system developed in such a way that the specific ‘this,’ which was presumably more marked than the middle member, was lost and the system became binary, with the proximal as the less marked. Prior to the Slavic invasions/migrations, a potential three-way opposition, however, was again on the rise, but the evidence is not clear enough to tell us how far it had gone by the time of the Slavic invasions/migrations, the fourth crusade, or the Ottoman conquest. Importantly, though, the evidence of these early modern grammars suggests that αὐτός did not yet occupy the place that it does today.
For South Slavic, in most of the region, the proximal demonstrative sǐ was eventually replaced by ovŭ (mutatis mutandis), which in OCS had a contrastive sense (Flier 1974: 59). The center of this innovation was the territory of the former Serbo-Croatian. It is found throughout Macedonian and in dialectal Slovene and Bulgarian (Vaillant 1958: 379–381). Pronominal s- is preserved as the proximal marker in the Balkan Slavic of the Rhodopes, and in the Korča region (Mazon 1936: 65–66; Steinke & Ylli 2007; Kokkas 2004; Stojkov 1993: 127; Adamou 2010b provides additional analysis of the opposition in the Pomak of Greek Thrace). For most of Bulgarian only a binary opposition remains, e.g., tozi/toja vs. onzi, while in Κorča Macedonian the binary opposition is s-/t-. For Slavic, the preservation of a three-way opposition with three etymologically distinct markers is found only in BSl plus non-Balkan BCMS.Footnote 110
Moreover, except in two isolated peripheral dialects – the central Rhodopes and northern Pindus – Slavic, like Greek, replaced its marked proximal demonstrative. Here, too, western Macedonian differs from eastern Macedonian and Bulgarian. In western Macedonian – as in former Serbo-Croatian – a restructured three-term system is preserved by replacing the old proximal with a pronoun that earlier meant ‘this one’ (ov-) in contrast with ‘the other one’ (on-).
In Macedonian, the oppositions are demonstratives ovoj/toj/onoj and locative adverbials ovde/tuka/tamo. Toj also competes with on as the third-person pronoun. In the demonstratives, ov- is proximal, on- is distal, and t- is neutral, whereas in the adverbials, ovde is speaker-oriented, tuka is addressee-oriented (which may or may not include speaker, e.g., when speaking on the telephone, the speaker uses tuka to ask if the sought person is located at the place to which s/he is calling), and tamo is distal. Thus, for example, when making a telephone call, in English one says ‘Is Ivan there?’ whereas in Macedonian one says Dali e Ivan tuka? (lit., ‘q is Ivan tuka’).
Balkan Slavic differs from the rest of South Slavic and from other Balkan languages in that there are five parameters of dialectal differentiation in deictics: (1) older three-term vs. newer two-term systems, (2) proximal/neutral vs. distal/neutral in two-term systems, (3) s vs. v as the proximal marker,Footnote 111 (4) absence vs. presence of postposed definite deixis marking, and (5) presence vs. absence of case marking on postposed articles. It should be clear from these parameters that the discussion here necessarily impinges somewhat on §6.1.2.2, where the definite article is given a fuller treatment. Table 6.12 summarizes the distribution and root consonant marker. A hyphen to the left indicates that the marker occurs in postposed definite articles; a hyphen to the right indicates that the marker can occur in freestanding deictics.
According to Mladenović 2001: 356, the loss of distal deictics in Lower Gora (from Mlike northward in Kosovo and all of Albania) occurs under the influence of Albanian, which, as indicated above, has the same type of binary deictic opposition. Prizren, like the rest of Kosovo and most of BCMS territory, has a three-way deictic distinction but no definite articles. Like Lower Gora, the Macedonian dialects of the Korça region also simplify to a proximal/neutral distinction, but with a different pronominal stem. Eastern Bulgarian, however, except in the Rhodopes and Thrace, has gone in the direction of East Slavic by losing the marker of proximity, but this change also occurs in the Kostur (Grk Kastoria) region, where Aromanian, Albanian, and, in the south, Greek, were all present as contact languages. Eastern Macedonian also has a two-way deictic system, like eastern Bulgarian, but, as in Lower Gora and Korça, it is the distal that is lost.Footnote 112 Finally, the Timok-Lužnica (and transitional northwestern Bulgarian) developments, like those of Albania, Kosovo, and the southern Rhodopes, show a preservation of case endings on the article. Although oblique substantives also occur elsewhere in western Macedonian, such marking never occurs on an article.
In terms of center/periphery relationships, the interactions between deixis and definiteness in dialectal Balkan Slavic display differing patterns of conservatism and innovation. The Korça and Rhodopian and Thracian developments are classic peripheral archaisms in their use of s- and their preservation of some case distinctions. The Rhodopian is more innovating in that s- is generally eliminated as a freestanding deictic, while in Thracian s- does not develop into an article. Gora and Timok-Lužnica share in the replacement of s- by v- but preserve case markers that western Macedonia does not. This leaves eastern Macedonia and Bulgaria, but also Kostur, as regions that have gone furthest in the direction of simplification. In terms of historical development, the grammaticalization of postposed deictics as articles must have involved all the deictics, and as the grammaticalization of definiteness proceeded, the marked deictic articles were gradually lost. The simplification in the western periphery is clearly contact-induced, but the eastern core is not as easy to explain. Nonetheless, given the relative stability of deictic systems in non-Slavic Balkan languages, the complexity of the Balkan Slavic situation indicates the importance of local conditions in systemic preservation and change.
The differential spread of the postposed definite articles and the reduplication of nonpronominal objects indicates that they mark different allegiances as well as different types of narrative strategy. The spread of the definite article to Gora and no further suggests its orientation to the south and the emblematic status of the definite article (cf. Enfield 2001: 267–268). Reduplication, on the other hand, is a different type of referentiality. The Prizren dialects, with their clear preservation of the accusative/dative opposition in substantives, argue against the case function of reduplication and in favor of its discourse function (see §6.1.2.5). At the same time, it appears to have become negatively valued in the southeast Serbian periphery during the course of the twentieth century.Footnote 113
Albanian displays only a binary opposition in the demonstratives, with a distinction based on the etymologically deictic particles a- (distal) and kë- (proximal). There are, though, three deictic adverbs of place – këtu ‘here,’ aty ‘yonder,’ and atje ‘there’;Footnote 114 Murzaku 2008 demonstrates the positive proximal markedness of këtu, the negative markedness of aty, and the neutrality of atje. Using collocational methodology, he shows how the deictic adverbs of direction (këndej/andej), manner (kështu/ashtu), and quantity (kaq/aq) correlate with the oppositions present/past, definite/indefinite, person/nonperson, venitive/andative, familiar/unfamiliar, known/unknown, and up/down, and how all map to varying degrees onto the proximal/distal opposition, which is basic for the deictic adverbs. He demonstrates that the three basic terms (këtu, aty, atje) form two sets of binary oppositions rather than one ternary one, much like the situation with Greek demonstrative pronouns or deictic adverbs. While këtu corresponds more to the personal in the deictic adverbs, it is the distal demonstrative that is the more “pronominal” of the two demonstratives. Murzaku also elucidates the problem of distinguishing the concepts of deictic pronoun from personal pronoun in the third person. He brings to bear the full power of corpus linguistics to demonstrate the fact – intuitive to the linguist who knows Albanian, but nonetheless heretofore unproven – that ai is more “personal” and ky is more “deictic” but that neither can be fully assigned to one or the other category.Footnote 115
There is less to say about the other languages. Balkan Romance seems to simply show the development found in most of Romance whereby the older three-way opposition of Latin (see above) has been reduced to a binary one, based on iste and ille. Thus one finds Rmn acest/ăst-acel/ăl, Aro aístu-atselu, and Megl tsistu-tsãl ‘that’ (Rosetti 1969: 247–252). Given the general Romance character of the Balkan Romance deictic system, the developments leading to the BRo system most likely predate any Balkan involvement.
Turkish in general seems simply to preserve an earlier three-way opposition with its bu, şu, o, and Balkan Turkish in particular shows no special developments of interest here.
Finally, Romani also has a binary opposition -da- ‘proximal’/-do- ‘distal.’ However, this opposition intersects with and combines with an opposition -ka-/-ko- which mark forms as specific (proximal/distal), so that every demonstrative must express both oppositions, giving a discourse-bound four-way opposition. Arli adava, odova, akava, okova (see Matras 2002: 103–106) exemplifies this unique quadripartite system. While there is considerable dialectal variation in the actual forms (see Matras 2002: 103–106), almost all Romani dialects both within and beyond the Balkans preserve such a system. Inasmuch as divergences among Balkan languages are as interesting and sometimes revealing as convergences, this Romani system is particularly noteworthy.
6.1.2.2 Definiteness
The issue of definiteness in the Balkans, especially regarding its specific realization as a postpositive article, is treated in various places in both earlier and later chapters, for different reasons. For instance, in Chapter 1, it provides a tiny window into a feature of a possible autochthonous prehistoric substrate language, through Hamp’s analysis of the ancient toponym Drobeta (see §1.2.1 and §6.1.2.2.1.2). And, as discussed in Chapter 2, it is an important part of the history of Balkan linguistics as a scholarly discipline, due to the notice of the postpositive article by early observers, including Leake (§2.1), Kopitar (§2.2.1), Schleicher (§2.2.2), and Miklosich (§2.2.3) in the nineteenth century, and by key early twentieth-century scholars such as Seliščev (§2.3.2.1) and Sandfeld (§2.3.2.2). Moreover, the history of the definite article within the Balkan languages and its geographic distribution make it an important and interesting case study regarding the methodology of argumentation within contact linguistics (see §3.2.2.5).Footnote 116 Finally, §7.4.1.1 treats some details about its ordering relative to other elements in the noun phrase, whereas §7.9.2 discusses definiteness in prepositional phrases.
Despite these mentions and especially the extensive treatment of the prehistory of the Balkan definite article in §3.2.2.5, there is still more to say about it, especially regarding its categorial status in Balkan grammars and its morphological and morphosyntactic realization and even about its prehistory. As it turns out, Leake’s 1814: 380 and Kopitar’s 1829 original recognition of the areal nature of the Balkan “postposed” definite article, given the time and place of its development, stands up to scrutiny and indeed reflects a Balkanism. Finally, for all the attention on the postposed articles, some attention to the preposed forms is in order too, especially with regard to Romani. We address these and related topics in the sections that follow.
6.1.2.2.1 Postposed Definiteness Marking
The presence of articles as markers of definiteness and indefiniteness within the noun phrase is a subject that cuts across a number of areas of analysis within Balkan grammar. They are at once expressions of morphosyntactic categories but also are elements analyzable as being at the borders of morphology and syntax. Indeed, for some of the languages, it is controversial whether these definite articles are clitics or affixes. Lunt 1952: 41 writes that Macedonian definite articles “are enclitics and can be termed suffixes,” while at one point Aronson 1968: 43 refers to the Bulgarian articles as desinences, but later (p. 46) writes: “The definite article is not a desinence, but an enclitic.” Halpern 1992: 194 calls the article an “inflection on the head of the first constituent of the NP.” Nicolae 2015 and Ledgeway 2017 argue that the postposed definite articles in Romanian are suffixes. The origin of the article obviously lies in the cliticization of a once-free deictic (see below), but some of its contemporary behavior is clearly affixal. Moreover, the answer to how they are to be categorized depends largely on one’s theoretical stance regarding affix-like/clitic-like phenomena and where to draw the line between them.Footnote 117
The elements of particular Balkanological interest, as noted in earlier chapters, are the definite articles in Balkan Slavic, Balkan Romance, and Albanian. They are generally referred to as “postposed” since they typically occur after the head noun in the noun phrase, as in (6.51):
a.
Romanian: om-ul ‘the man’ man-the b.
Aromanian: om-lu ‘the man’ man-the c.
Meglenoromanian: om-u ‘the’man’ man-the d.
Macedonian: grad-ot ‘the city’ city-the e.
Bulgarian: grad-ăt ‘the city’ city-the f.
Torlak BCMS:Footnote 118 brat-at ‘the brother’ brother-the g.
Albanian: kal-i ‘the horse’ horse-the
By contrast, Greek and Romani here have preposed articles: Grk ο άνθρωπος ‘the man’ (*άνθρωπος ο), Rmi o manuš (*manuš o) ‘idem.’Footnote 119
Given these facts, it is indeed fair to describe the article as postposed relative to the noun, and even fuller noun phrases, with modifiers, give that impression:Footnote 120
a.
Romanian: om-ul bun ‘the good man’ man-the good b.
Aromanian: om-lu bun ‘the good man’ man-the good c.
Meglenoromanian om-u bun ‘the good man’ man-the good d.
Macedonian: grad-ot nov ‘the new city’ (marked order and effect) city-the new e.
Bulgarian: grad-ăt nov ‘the new city’ (marked order and effect) city-the new f.
Albanian: kal-i i madh ‘the big horse’ horse-the PC big
However, in Balkan Slavic it is normal – and in the others at least possible – for the adjective to occupy initial position in the noun phrase, and in that case, the definite article attaches to the adjective and not to the noun:Footnote 121
a.
Romanian: bun-ul om ‘the good man’ / *bun om-ul good-the man b.
Aromanian: bun-ul om ‘the good man’ / *bun om-ul good-the man c.
Meglenoromanian bun-u om ‘the good man’ / *bun om-u good-the man d.
Macedonian: novi-ot grad ‘the new city’ / *nov grad-ot new-the city e.
Bulgarian: novi-jăt grad ‘the new city’ / *nov grad-ăt new-the city f.
Torlak BCMS: četvrto-to jutro ‘the fourth morning’ fourth-the morning g.
Albanian: i madh-i kal ‘the big horse’ / *i madh kal-i pc big-the horse
What these facts show is that the definite article is positioned relative to the elements making up the noun phrase, not to the noun proper; that is, as Halpern’s formulation cited above suggests, the article is in second constituent position in the noun phrase, occurring after the first inflectable element, whatever that may be.Footnote 122 As noted in §5.3, this sort of postpositive (enclitic) behavior, involving occurrence in second position relative to some domain, is quite familiar cross-linguistically with prosodically deficient and dependent items, and is generally referred to as Wackernagel positioning or labelled as an instance of Wackernagel’s Law, after the formulation by Jacob Wackernagel 1892 concerning the behavior of unaccented forms in ancient Indo-European languages.
This fact of second constituent position is a crucial aspect differentiating the Balkan “postposed” definite marker from other postposed manifestations of definiteness marking in related languages, specifically in Armenian and Scandinavian languages, with North Russian dialects sometimes also included. In those languages, the definite article occurs after a noun, parallel to the facts in (6.51), but is not “postposed” in the Balkan fashion. Rather, in those languages, unlike the Balkan situation seen in (6.53), the article attaches only to a substantive and never to a preposed adjective. For instance, Armenian has mart-ə ‘man-the’ for ‘the man,’ but lav mart-ə ‘good man-the’ for ‘the good man,’ and Norwegian has for ‘the old horse’ den gamle hesten, literally ‘the old horse.def,’ with a determiner (den) and definiteness marking on the noun (hest-en). Moreover, in Scandinavian, the postposed article is in competition with a preposed article,Footnote 123 while in North Russian it is actually generally analyzed as a demonstrative or emphatic particle or topicalization marker rather than as an article (Seliščev 1968: 200–220; cf. also Koduhov 1953; Wiemer & Hansen 2012: 114–116 on topicalization).
The historical background to the array of Balkan facts is both complex and interesting. Accordingly, the facts from each of the languages are taken up in turn in the following sections.
6.1.2.2.1.1 Balkan Slavic
The one language group for which we have thorough attestation on the phases and complexities in the development of the Balkan-style definite article is Slavic, and there is no question that it began after the arrival of the Slavs in the Balkans and reached its current shape in the Ottoman era.
Still, it must be noted first that in Slavic, taking OCS as representative of Common Slavic, definiteness was marked on adjectives by the suffixation of the old relative and subsequently 3sg pronoun *jь. This innovation most likely dates to the Balto-Slavic period, as it is seen in Lithuanian, e.g., gẽras – gerȧsis, equivalent to OCS dobrъ – dobrъjь ‘good’ – ‘the good’ (Fortson 2010: 367). The antiquity of this development is suggestive vis-à-vis Albanian. If, as Hamp 1991, 1992b, 2010 argues (on the basis of ancient shared innovations and borrowings), the Indo-European dialect that became Albanian and the one that became Slavic were in contact prior to the migration of the speakers of Albanoid to the Balkans, then the postposing of definiteness marking might have been a shared tendency.
This marking on adjectives, giving so-called long-form adjectives, survives morphologically to one degree or another in all the Slavic languages, but the meaning of definiteness associated with it is found only in its remnants in BCMS and Balkan Slavic. In Macedonian, for instance, in the pattern of (6.53), with the article attaching to the adjective, the form of the adjective, e.g., novi-, reflects an old long form that is otherwise quite restricted synchronically, occurring just in some toponyms, e.g., Gorni Saraj ‘Upper Saray’ (a neighborhood in Ohrid), in folk poetry, and in the vocative expressions of the type dragi moj [prijatele] ‘O my dear [friend],’ etc. Further as to deixis and definiteness in Slavic, in much of the Slavic periphery, the neutral deictic in t- has come to mark definiteness. It appears as a preposed deictic in Czech and Slovak and Slovene (and in some colloquial Polish) – all languages in contact with German (or, in the case of Slovak, Hungarian) and thus with preposed articles – and as a postposed element in North Russian dialects (Kuznecov & Bromlej 1973; but see §6.1.2.2.1), perhaps under Scandinavian or Uralic influence.Footnote 124 But in all the regions the usage is colloquial/dialectal and somewhat marginal or facultative, or pragmatically determined, and in none of them is it as systematically integrated into the grammar as in Balkan Slavic. Moreover, only in Balkan Slavic does definiteness intersect with differential deixis; that is, Torlak BCMS, western Macedonian, and Rhodopian and Thracian Bulgarian all have articles derived from demonstratives in addition to (or other than) t-, especially v- or s- and n-, allowing for deictic contrasts under the rubric of definiteness. The maps in Mladenova 2007: 433–436 show the tripartite system being used in Bulgarian in south-central Bulgaria along the border of Greece as well as in the area around Trăn on the border of Serbia, in western North Macedonia and into Kosovo. That the peripheral systems are both more archaic and more complex, however, suggests that simplification of a three-way system could have been the innovation, rather than elaboration of a one-dimensional definitivizing system.
Another key point in the spatial distribution of definiteness marking in Slavic is that while the non-Balkan BCMS dialects as well as Slovene preserve remnants of *jь-derived definiteness marking on the adjective, there is no deployment of t- as a grammatical definiteness marker.Footnote 125 In fact, the postposed definite article of eastern Torlak is a salient marker of those dialects and is perceived as an object of humorous imitation by speakers from other areas. Thus for BCMS, the postposed definite article is emblematic of southernness, i.e., Balkanness. Mladenova 2007: 344–367, by identifying NE Bulgaria as a nucleus of innovation, is pointing to precisely the region that was most in contact with Romance and also most subject to influence from Macedonian and Albanian speakers who migrated there from the southwest during the Ottoman period. As Mladenova’s research makes clear, this was the period when the BSl definite systems as we know them today took their definitive shape, although their beginnings are earlier.
As for the positioning of the deictic element, already in the earliest Old Church Slavonic manuscripts, the demonstratives sъ ‘this’ and tь ‘that’ can occur postposed. Moreover, these postposed demonstratives cliticized onto their host nouns, as indicated by the vocalization of what would otherwise be a final (weak) jer, e.g., plodъ – plodosь ‘fruit – this fruit,’ domъ – domotъ ‘house – that house,’ denь – denesь ‘day – this day’ (Diels 1963: 154, 163). Although such vocalization is by no means regular in the manuscripts, the fact that it occurs at all is an indication that, at least from a phonological viewpoint, the process that would lead to the postposed definite article was already underway when these manuscripts were written, presumably in the tenth or eleventh century.
One hypothesis on the emergence of a definite article in Balkan Slavic sees pre-Ottoman Turkic, specifically Bulgar Turkic (Proto-Bulgar), as playing a key role. Mladenova 2007: 357–362, citing Kusmenko 2003, resuscitates and elaborates this argument, basing the claim largely on the fact that Turkic languages all have definite object marking and so-called personal declension (i.e., possessive suffixes), of which the third person marker interacts with definiteness. Kusmenko 2003: 145, moreover, compares the person-based three-way deictic system of Turkic with the three-article system of BSl. However, this account is less than convincing. Aside from the fact that the tripartite article system occurs in areas far removed from pre-Ottoman Turkic contact (Rhodopes, Thrace, Gora, Torlakia), there is no evidence whatsoever that Albanian ever had such a system, and, perhaps most importantly, no such evidence for Balkan Romance. Moreover, the system in Balkan Slavic is not really person based. It is, in fact speaker oriented and expresses the speaker’s attitude, e.g., Macedonian ezero-vo ‘lake-def.prox’ when an Ohrid speaker refers to Lake Ohrid to a fellow Ohrid speaker, but ezero-no ‘lake-def.dist’ when addressing a person from Skopje.Footnote 126 Another important point of difference between the Balkan and the Turkic definite marking (as also between Balkan and Scandinavian, North Russian, and Armenian – see above in §6.1.2.2.1) is the second-position placement of the article. In particular, the definite accusative in Turkish that would correspond to, say, Macedonian ubaviot čovek ‘the handsome person’ is güzel adamı and not *güzeli adam. It is hard, therefore, to see how contact with Turkic definite usage could lead to the Balkan Slavic definite article, given the differences in their positional behavior. Moreover, Turkic word order behaves differently from that of Balkan Slavic, Balkan Romance, and Albanian.
If language contact is a possible source of the Slavic definite marking, then one might well look to Balkan Romance, as Sandfeld 1930: 170–171 suggested (see §6.1.2.2.1.3), yet the role of Balkan Romance is not without controversy. Gălăbov 1962: 115ff., in denying the influence of Balkan Romance on Balkan Slavic article formation, makes the point that article usage in Bulgarian is considerably more limited in its various types of usages than in Romanian. However, one can just as easily use that difference to argue in favor of a Balkan Romance (or, more probably, Balkan Latin) influence, since precisely that lesser complexity points to the kind of simplification that sometimes takes place in contact situations. A fuller account of the source of the Balkan Slavic article is deferred until §6.1.2.2.1.5, but Balkan Romance is argued to have had a role in the development of the article, along with other factors (Gołąb 1997).
6.1.2.2.1.2 Albanian
Key issues in the examination of definite marking in Albanian are not just its ultimate source but also how it came to be postposed in the way it is. Complicating the matter, but making it more interesting at the same time, is the issue of the Albanian particle of concord, a determiner-like bit of morphosyntax having the shape i, e, të, or së that is relevant here because it has historical connections with the definite article, but also because it has an interesting parallel in Balkan Romance.Footnote 127 It serves mostly a connective function, preceding genitives and certain adjectives when they modify a noun, but also occurs with those adjectives when they are used predicatively:Footnote 128
a.
Djali i mirë boy.def.nom pc good ‘the good boy’ b.
Libri i Agimit book.def.nom pc Agim.gen ‘the book of Agim’ / ‘Agim’s book’ c.
e gjeti mësuesin e lumtur it.acc found.3sg teacher.def.acc pc happy ‘He found the happy teacher’ (Newmark et al. 1980: 124) d.
e kishin bërë rojtar të fshatit him.acc had.3pl made sheriff pc village.gen ‘They had made him sheriff of the village’ (Newmark et al. 1982: 126)
The particles of concord are clearly of the same demonstrative origin as the postposed marking for definiteness, as the two share much in form; as Newmark et al. 1982: 179 put it:
These preposed articles are connected historically with postposed articles which lost their independence and turned into definite case endings. Note the similarity of form of the proclitics [i.e., the particles of concord] and the definite case endings for the same word in the following examples: I mir-I ‘the good one (m),’ TË mirë-T ‘the good ones (m),’ SË mirë-S ‘to the good one (f)’ TË miri-T ‘to the good one (m).’
There is debate as to how Albanian could have both preposed and postposed (historical) demonstratives serving somewhat different functions and which position is older and (more) original.
There are two competing schools of thought here: one holds that definite marking began as enclisis out of which proclisis developed (e.g., Çabej 1963, 1979; Hamp 1982; Sh. Demiraj 1985), the other is that it began as proclisis and enclisis followed (e.g., Camarda, Pedersen, Riza, Bokshi). K. Topalli 2009 gives a useful summary of the main references and of Çabej’s arguments, and Morina 2012 focuses specifically on Bokshi vs. Çabej in this matter. There are also arguments concerning its dating, but basic phonetic facts such as the preservation of stem-final nasals (sŷ ‘eye,’ def syni/syri Geg/Tosk) place the formation after the loss of final consonants inherited from Indo-European but before the Geg/Tosk split, i.e., prior to the arrival of the Slavs (sixth century) but not necessarily prior to the arrival of the Romans.
A relevant piece of the picture is the form of the definite accusative ending, -në, as in nusenë (StAlb nusen), from nuse ‘bride, daughter-in-law.’ Çabej 1963: 74, citing Pedersen, gives a reconstruction *snusām-tām, with the inherited accusative of *snusā followed by a (postposed) demonstrative; the path of development would have been mt > nt > nd > n (presumably via -nn-), with the outcome of *nt seen in the 3pl ending -(a/j)në from *-nti (as in janë ‘they are,’ ultimately from *H1senti). This process pre-dates the period of Latin influence on Albanian, since Latin loanwords with -nt- preserve and voice the stop, as in cantō ‘I sing’ > këndoj, centum ‘hundred’ > qind.Footnote 129 The analysis in Hamp 1982: 79 of the toponym Drobeta is relevant here. As noted in §1.2.1, he saw a form *druṷā tā, literally ‘wooded(-place) the,’ as underlying Drobeta; this form, with what he posits as a postposed definitizing element (*tā) and belonging to an Indo-European Balkan substrate language, was misinterpreted by Romans encountering this name.Footnote 130 This account, if accepted, gives added credibility to Pedersen’s reconstruction, and moreover allows for the possibility that the Albanian postposed (enclitic) definite article derives from the pattern in the language whose form *druṷā-tā the Romans misanalyzed (see §6.1.2.2.1.5 for more discussion).
Sh. Demiraj 1985: 297–357, 1057–1061; 2002: 151–180 gives a thorough and balanced account of all the arguments on various sides, but ultimately agrees with Hamp (and Çabej, as cited in Sh. Demiraj 2002: 161). Asenova 2002: 127 notes that the preposed particle of concord is less frequent and more restricted, and that moreover, when it serves a linking function with adjectives and genitives its shape is determined nonetheless by what precedes it; that is to say, even when “preposed,” it is syntagmatically enclitic, an argument for an origin in enclisis. In such a scenario, the preposed forms that are more independent and do not serve a connective function would have been the result of the reanalysis of the marker as properly belonging to the adjective and being generalized from there. The Albanian postpositive definiteness marking would thus ultimately have its origin in the substrate pattern seen in Drobeta, whatever language (or predecessor to a modern language) that substrate may have been.Footnote 131
As noted above, there is a parallel in the Balkans to the Albanian particle of concord, namely the determiner cel in Balkan Romance. There is Balkanological significance to this Romance form and to this parallel, discussion of which is presented in the next section.
6.1.2.2.1.3 Balkan Romance
An account of postpositive definiteness marking in Balkan Romance must start with an understanding of the situation in Latin, even if the eventual system turns out not to be an inheritance per se. As Gołąb 1997 notes, the Latin distal deictic ille was generalized as an unmarked pronoun/deictic/article, as was the case in most of Romance (except Sardinia, Majorca, and parts of Gascony and Catalonia, where ipse was used), but, unlike any Romance language outside the Balkans, Balkan Romance definite articles are postposed.Footnote 132 Postposed deixis was an acceptable strategy in Latin – as in Old Church Slavonic – a fact which has led some scholars to view the development as purely internal (e.g., Mihăescu 1978: 257 and references therein, cf. also Maiden et al. 2021: 202), with a postposed article being just another type of postposed modification of a noun; schematically, this development can be represented thus: det-adj noun > noun det-adj > noun-det (adj). Stan 2013b: 286, citing Coteanu 1969: 96–97 and Philippide 2011: 460, sums up this internal-evolution account as follows:
On the basis of the primitive structure homo ille-bonus [‘man that-one good’] or homo ille + genitive, a continuation of the Latin pattern homo ille (Bourciez 1956: 248), where ille, originally associated with the adjective / genitive, as a DP-internal constituent, was reanalyzed and grouped with the noun, in postposition.
Such a view, however, while certainly plausible,Footnote 133 forces one to assume that Balkan Romance and Balkan Slavic, which we know from other evidence were in intensive contact (loanwords, politics, etc., cf. especially Gołąb 1970, 1976, 1997), just happened to develop postposed enclitic definite articles at the same time and in contradistinction to developments in the rest of Romance and Slavic, respectively.Footnote 134 One way around this dilemma is to suggest, as Sandfeld 1930: 170 did, that the Slavic situation was due to contact with Romanian (i.e., Balkan Romance). As for the question of what to do about the parallel development in Albanian, Sandfeld’s answer was that the seeds of article development could have been a part of Albanian and Balkan Latin from early on, and that Albanian speakers could have been influenced by Balkan Latin, which could well have had “certaines tendances à la postposition” (‘certain tendencies to postposing’).
There is a lot of speculation in this account, though it does draw on the known quantity of the availability of preposing and postposing for demonstratives in Vulgar Latin, and early contact between Albanian speakers and Balkan Latin speakers is a given, based on the large number of Latin loanwords in Albanian dating from the Roman period in the Balkans (see §4.2.1.2). Moreover, contact between Romanian and Bulgarian, at least, is also a given, though it is not clear that the chronology of known contact is right for what Sandfeld wanted to explain.
Still, there is a better account, one that exploits the Albanian–Romance connection and the Romance–Slavic connection. This is the account based on Hamp 1982 and his analysis of the place name Drobeta, outlined in §6.1.2.2.1.2. His idea that a substrate language, as seen in the form underlying Drobeta, was an Indo-European language with postpositive definiteness marking (with *tā in the feminine, no less) raises the possibility that Balkan Romance acquired that feature as substrate speakers shifted to (Balkan) Latin. In that shift, they could have opted for the postpositive possibility that Latin offered at that time for demonstratives, inasmuch as it matched their indigenous syntagma. To the extent that there might be a relationship between Hamp’s substrate language, what he artfully calls simply an “autochthonous language,” and the language that became Albanian, this account allows one to exploit the evidence of a very old connection between Balkan Romance and Albanian evident from a number of loanwords shared just by those two Balkan branches (see §4.2.1.1) and the possibility of a shared substrate that this vocabulary might point to.
Such a substrate may also prove useful – and taking Drobeta as overt evidence, less speculative than most substratum accounts of prehistoric linguistic phenomena – when it comes to understanding the parallel between Albanian and Balkan Romance with regard to the presence of an adjectival article. This is the particle of concord discussed for Albanian in §6.1.2.2.1.2, for which the Balkan Romance parallel is the determiner cel. This form “functions both as a freestanding definite article and as an adjectival article,” according to Nicolae 2013: 309, who gives the following examples:
a.
băiatul cel mare boy.def CEL big ‘the big boy’ b.
cel mare CEL big ‘the big one’
A similar usage can be found in some varieties of Aromanian with the ostensible demonstrative atsel-; Capidan 1932: 400 cites phrases like that in (6.56a) for the Aromanian of Greece, and more recently, Campos 2005: 313 gives (6.56b) for what he calls Arvanitovlaxika, “a dialect of Aromanian spoken in Thessaly, Greece” (in modernized orthography):
a.
om-lu atsel bun-(lu) man-the CEL good-(the) ‘the good man’ b.
pulji-lji atselji njits-lji birds-the “those” small-the ‘the small birds’
And in the Aromanian of North Macedonia, while the use of a form like tsel is rare at best, one usage is evident, namely cal with the first name of a man to name his wife, e.g., husband Dona has a wife Caldona; this is a connective to an understood head noun, i.e., ‘(wife) of/associated-with Dona.’ This usage can be interpreted as a remnant of an earlier broader use of such a connective element. Moreover, it is found in Meglenoromanian as described by Capidan 1925a: 150, e.g., fitšoru ţela bun ‘the good son.’ The occurrence of a cel-related form in a similar function in Aromanian, Meglenoromanian, and Romanian means the entry of the construction into Balkan Romance must be placed prior to the divergence of Aromanian and Meglenoromanian from Romanian, and thus relatively early in the emergence of Balkan Romance from Late Latin.
Romanian cel and the Albanian particle of concord, as a comparison of (6.54a) with (6.55a) shows, are thus similar as to their placement relative to the noun and modifier and as to their functions, with both showing a linking function; they also both have a nominalizing function, as a comparison of (6.55b) with Albanian participles nominalized with a particle of concord shows, e.g., të folurit ‘speaking.’ Moreover, both are similar as to their origin, in that cel derives from “the endophoric distal demonstrative acel / acela (< ecce / eccum + illum)” (Nicolae 2013: 310), like the form seen in Aromanian, just as the Albanian particle of concord has a demonstrative origin. Still, there are some real differences: the Balkan Romance cel is basically optional, whereas the particle of concord, while lexically optional in that there are some adjectives that take it and some that do not, is obligatory with those adjectives that take it; Romanian cel has a definitivizing function – Nicolae ibid. speaks of (6.55a) as being “a double definite construction, since definiteness is expressed twice (by the affixal definite article and by the adjectival article)” – whereas the Albanian particle of concord is not connected with definiteness and can occur in indefinite noun phrases, e.g., djalë i mirë ‘a good boy’ (lit., ‘boy.indf pc good’), i një djali të mirë ‘of a good boy’ (lit., ‘pc one boy.gen.indf pc.gen.m.indf good’). Still, the parallelism is striking and may well reflect an old substratum feature shaping the internal syntax of the noun phrase in both Albanian and Balkan Romance, if the “Drobeta language” had both the attested postpositive article and an adjectival article, related formally.
Such a conclusion is not without problems, though they are not insurmountable. First, it must be admitted that if one focuses on the differences between the Albanian and Balkan Romance uses, rather than the similarities, an alternative conclusion might well be reached; still, the similarities are not to be denied. Second, this account is at odds with the scenario given in §6.1.2.2.1.2 for the development of the particle of concord in purely Albanian-internal terms, and with the Balkan Romance-internal account given by Nicolae 2013: 310 (citing others as well) for cel; this, however, is not a serious problem, as the internal accounts need not be right, even if their inherent plausibility gives them some credibility from the start. Third, there may be chronological issues, as Iordan & Manoliu 1965: 145 claim that cel arose only after the sixteenth century; here, though, one must bear in mind that Romanian per se is only attested from the sixteenth century and in any case, others have a different take on the situation. Iliescu 2006, for instance, sees the developments leading to cel as part of a long cycle of deictic expressivity found across Romance, and thus from Late Latin on, with proclitic (preposed) elements; indeed, as noted above, the parallel between Romanian and Aromanian on the use of the adjectival article dates it as a relatively early structural aspect for the languages.
One further detail about the “double definite” construction is of Balkanological interest and importance. Campos 2005: 318, noting the Aromanian structure in (6.56) with its multiple overt expression of a definite article (-lu in (6.56a), -l’i in (6.56b)), beyond the definiteness inherent in the CEL form, points to a parallel structure in Greek:
(6.57)
το πουλί (αυτό) το μικρό the.sg bird.sg this.sg the.sg small.sg ‘the small bird’
Campos claims that the Aromanian structure of (6.56) is borrowed from Greek, which he says has a parallel phrasing. However, it is not clear that (6.57) has the intended reading; some speakers report that (6.57) has only a reading with clear demonstrative force, ‘this small bird,’ a reading not available in Aromanian, rather than the definite article reading. This suggests that the historical relationship between the Greek and the Aromanian here may be tenuous at best. Indeed, it may well instead be the case that the Greek construction is borrowed from the Aromanian, not the reverse that Campos assumes.Footnote 135
To return to the question of the relationship, if any, between Balkan Romance as realized in Romanian cel and Albanian particles of concord, as with almost any substratum account, the suggestion of a connection cannot be definitively proved. The problems adduced, even though they can be countered, may doom it in some scholars’ minds. But, importantly, it is consistent with other facts that suggest an ancient relationship between Albanian and Balkan Romance, and with the compelling analysis of Drobeta offered by Hamp. Thus, it cannot be dismissed out of hand and needs to be given serious consideration in a Balkanological context.
One final point to be made about Romance languages in the Balkans and definite articles is that the one Romance language not considered under the rubric of Balkan Romance, namely Judezmo, does not show the Balkan-style postpositive article. This is not surprising, however, and needs no special explanation. Given the fact that by the time Judezmo entered the Balkans it already had a well-developed definite article system, with the prenominal positioning characteristic of Ibero-Romance, one would not expect a positional change to have occurred; the speakers simply retained the position for the article that they came to the Balkans with.
6.1.2.2.1.4 Where Greek Does, or Does Not, Fit In
As the coda to (6.51) above in §6.1.2.2.1 shows, Greek does not have a postposed definite article of the sort seen in Balkan Slavic, Albanian, or Balkan Romance. Rather, it has a definite article that, for as long as it has been a part of Greek, occurs before the element it combines with. The article developed within historical Greek, being absent from Mycenaean Greek and mostly from Homeric Greek, and coming into being fully only by the time of Classical Greek. The source was the Indo-European deictic pronoun *so/sā/tod (m/f/n, cf. Sanskrit sa/sā /tad, Gothic sa/sō /þat-a), which gave Greek ὁ/ἡ/τό.
Article use has remained a part of Greek nominal syntax ever since then, throughout the history of the language, yet most handbook-style accounts (e.g., Schaller 1975, as a representative) of the development of the articles in Albanian, Balkan Slavic, and Balkan Romance, list Greek as not showing any relevant “Balkan” article characteristics. This is largely due to the fact that the Greek article is not a postposed enclitic element within the noun phrase. Moreover, the Greek article shows various characteristics that set it off from articles in other of the standard Balkan languages, features that are all, as it turns out, survivals of Classical Greek usage. But there are other traits that show potential Balkanological significance upon closer inspection. Still, even ones that are distinctly Greek are of interest once it is recognized, as argued in the Introduction to this book and in §3.3, that it is valuable to consider divergences among the Balkan languages in the midst of so much convergence, as they shed some light on the dynamics, the give-and-take, of the contact situation, on what is transferred across languages and what is not.
The first of these is a usage that is unique, or nearly so, in the Balkans, namely the nominalizing use of article. The article can be preposed to any constituent to give a nominal. Most notable is the use of the neuter form to nominalize clauses, as in (6.58):
a.
δεν μου αρέσει το πώς με κοιτάζει neg me.gen please.prs.3sg the.n how me.acc look.at.prs.3sg ‘I don’t like how he looks at me’ (lit., ‘ … the how he looks at me’) b.
το ότι είναι ένοχος είναι το πρόβλημα the.n comp be.prs.3sg guilty be.prs.3sg the.sg problem.sg ‘(The fact) That he is guilty is the problem’ (lit., ‘the that he is guilty … ’) c.
το να αργήσουμε τον ενοχλεί the.n dms be.late.pfv.1pl him.acc bother.prs.3sg ‘Our being late bothers him’ (lit., ‘the that we-are-late … ’)
In this form, a clause can be the object of a preposition, as in (6.59a) and the article can even be inflected, as in (6.59b), where it is in the genitive case:
a.
με το να στέκομαι εκεί δεν βολεύτηκε ο Γιάννης with the.n dms stand.prs.1sg there neg be.comfortable.pst.3sg the.nom Janis.nom ‘With me/my standing there, Janis was not comfortable’ (lit., ‘with the that … ’) b.
η θέα του να πιάνει τον Γιάννη the.nom sight.nom the.n.gen dms grab.prs.3sg the.acc Janis.acc η αστυνομία με τρόμαξε the.nom police.nom me.acc scare.pst.3sg ‘The sight of the police grabbing Janis scared me’ (lit., ‘the sight of the that … ’)
This usage continues an Ancient Greek use of the article, e.g., (6.58c) is the modern version of the ancient infinitival nominalization, the so-called “articular infinitive.” The closest that any other Balkan language comes to this is the use of the article with a verbal noun in Balkan Slavic, e.g., Macedonian prašenjeto ‘the (act of) asking,’ or the Romanian definite form of the long infinitive in -re, e.g., mergerea ‘the (act of) going’; however, those represent the use of the definite article on a lexicalized (or synchronically still actively derived) deverbal nominal, so that article is not attaching to a clause per se to nominalize it but rather is definitizing an already nominalized form.
As for the other properties, Modern Greek shows double determination (see also §6.1.2.3) with demonstratives, the possibility of doubled article use – polydefiniteness – in noun phrases when an adjective accompanies the noun, and special placement with the modifier ‘all.’ In each case, the modern state of affairs, even though continuing an ancient usage and not necessarily found in all of the other standard languages of the Balkans, nonetheless in certain respects situates Greek as more Balkan-like in article use than is generally admitted in the literature.
For instance, although prosodically weak in that some forms carry no accent, the Greek definite article, as noted above, is positionally restricted differently from the article in the other languages. It can, and typically must, occur as the first element in the noun phrase, and not the second element, as shown in (6.60):
a. το κόκκινο κρασί ‘the red wine’
b. *κόκκινο το κρασί
c. *κρασί το κόκκινο
d. *κρασί κόκκινο το
However, there are two circumstances in which the article appears in second position: with a demonstrative (αυτός ‘this,’ εκείνος ‘that,’ τούτος ‘that here’) and with όλος ‘all,’ as in (6.61):
a.
αυτός ο άνθρωπος / *ο αυτός άνθρωπος this.nom.sg.m the.nom.sg.m man.nom.sg.m ‘this man’ b.
όλος ο κόσμος / *ο όλος κόσμος all.nom.sg.m the.nom.sg.m world.nom.sg.m ‘all the world’
The usage with the demonstrative is a type of double determination, a feature found in other Balkan languages and arguably a Balkanism, as discussed more fully in §6.1.2.3. Other than these second-position uses, however, the Greek article is initial in the noun phrase, even when there is a large amount of material separating it from the head noun:
(6.62)
ο καλός τίμιος άνθρωπος the.nom.sg.m good.nom.sg.m honest.nom.sg.m man.nom.sg.m ‘the good [and] honest man’
The separation of the article and noun from one another by other words, as in (6.62), is a reason for treating the article as a word in itself, following what Dai 1991, 1992: 32–37, drawing on Bloomfield 1933, has called the “expansion test” for wordhood; in fact, a rather large amount of material, with internal structure of its own, can occur between the article and the head noun, giving noun phrases with preposed “extended modifiers,” as in (6.63):
(6.63)
ο μορφωμένος στο Παρίσι γείτονάς μου the.nom.sg.m educated.nom.sg.m in.the Paris.acc.sg neighbor.nom.sg.m me.gen ‘my educated-in-Paris neighbor’ (i.e., ‘my neighbor who was educated in Paris’)
This trait continues an ancient pattern of extended pre-nominal modifiers being allowable within (i.e., to the right of) the article; Smyth (1920: §1164), for instance, cites the following, where six words representing three different constituents intervene between the article and the noun it is modifying:
(6.64)
πρὸς τὴν ἐκ τῆς Σικελίᾱς τῶν Ἀθηναίων μεγάλην κακοπρᾱγίᾱν to the.acc.sg from the.gen Sicily.gen the.gen.pl Athenians.gen great.acc failure.acc ‘with regard to the great failure of the Athenians in Sicily’ (lit., ‘to the from the Sicily of the Athenians great failure’)
Thus if the article is a clitic word of any sort in Greek, it would have to be considered a proclitic, and not an enclitic, and therefore somewhat different in its behavior from the Balkan postposed article.
To return to (6.61b), the article’s occurrence in second position with ‘all’ όλος is potentially of Balkanological interest. In Classical Greek, there was a distinction made with ὅλος, based on the positioning of the article. It was a subtle distinction, between the entity taken as a whole and all of the pieces of the entity together. Smyth (1920: §1175) exemplifies it with στράτευμα ‘army’: when ὅλος is within (to the right of) the article, i.e., τὸ ὅλον στράτευμα, the phrase means ‘the whole army’ in the sense of all of its pieces (soldiers) being taken into consideration, whereas when ὅλος is outside of the article, i.e., ὅλον τὸ στράτευμα or τὸ στράτευμα ὅλον, the meaning is ‘the army as a whole.’ What is Balkanologially interesting is that in Modern Greek, the only positioning that is possible is the second possibility, outside of the article and thus with the article surrounded by όλος and the head noun: όλος ο κόσμος ‘all the world, everyone.’ That is, in the competition between placement of ‘all’ relative to the article, a Balkan-style positioning for the article has won out in this one phrase. It is hard to say if this article placement is the result of contact, but the change in possibilities from Ancient Greek is indisputable as is the fact that it is in the direction of the Balkan post-modifier position within the noun phrase. This may well be a case where Greek is a recipient of contact-induced pressure for change from other Balkan languages.Footnote 136 It can be noted too that the double determination with demonstratives as given in (6.61a) is also a case of Balkan-style article positioning, although in this case there was no alternative placement in Classical Greek; it has been postpositive placement of the article with demonstratives throughout the history of Greek with no deviations.
One other context in which Greek has a definite article occurring in an NP-interior position is in the polydefinite construction in which the article is repeated with a modifier after the head noun. An example is given in (6.65):
(6.65)
το κρασί το κόκκινο the.sg wine.sg the.sg red.sg ‘the red wine’
This continues a Classical Greek construction (e.g., ὁ ἀνήρ ὁ σοφός ‘the wise man’ (lit., ‘the man the wise’)), and has generated a considerable amount of literature, mostly aimed at the details of its proper analysis within Modern Greek.Footnote 137 Some scholars, e.g., Schaller 1975: 144, have noted the existence of this construction and have compared it to some syntax of articular origin found elsewhere in the Balkans with what from an etymological standpoint are multiple demonstratives; that is, the “copulative” article in Albanian and Romanian in which an adjective is linked to a noun by an overt connective (what Nicolae 2013: 309 for the Romanian cel construction calls a “double definite construction”).Footnote 138 And the postpositive nature of the repeated article is suggestive of the postposed definite article more generally. It is certainly the case that with the position of the second article in (6.65), the phrase appears as Noun-Article-Adjective, just as in (6.52) above. But a key difference between (6.65) and (6.52) is the prosodic fact that the article in Greek is not enclitic within the noun phrase and is not enclitic with regard to the head noun. Moreover, the polydefinite construction can have the adjective, with the article, before the noun (το κόκκινο το κρασί). Thus the key feature of the Balkan postpositive definite marking is lacking even in this superficially similar construction, so that looking to Greek as a source for the Balkan article developments is missing a crucial element. Admittedly, one could argue that the positioning is what was borrowed and the enclitic behavior is a language-internal development in each case. But a problem here is that this Greek construction, while a possible one, was less common (so Smyth 1920: §1158) in Classical Greek than the construction with a single article (ὁ σοφός ἀνήρ ‘the wise man’); further, another possible construction had a single article after the noun and before the adjective (ἀνήρ ὁ σοφός ‘man the wise’) but it was the least common of all (Smyth 1920: §1159), and in some instances was more of an afterthought (i.e., ‘a man, I mean the wise one’). It seems rather unlikely for somewhat marginal constructions to be the basis for diffusion via contact, when one might instead expect the predominant construction to be the one that second-language speakers would pick up on.
Thus the assessment of Greek as marginal at best within the Balkans based on the behavior of definite marking is appropriate, but it is not irrelevant to a consideration of the key Balkan trait. Greek shows interesting similarities and differences to what other Balkan languages show in this domain, but it is lacking in the key structural details that the other languages show. This is not a surprising situation, given that Greek already had a (generally preposed) definite article before the arrival of the Romans and the Slavs, and before the period of key Balkan contact; in this way, Greek is not unlike Judezmo, as discussed in §6.1.2.2.1.3.
6.1.2.2.1.5 Summation, and an Origin Story
To some extent, at issue in trying to understand the origins of the postposed definite article are questions of time depth. The basic facts as seen above are these:
(a) Greek developed a preposed definite article of deictic origin during the period between Mycenaean and Classical literacy, although it was not yet fully established at the time of the composition of the Homeric epics (c. 800 BCE).
(b) The language that became Albanian developed a postposed definite article from the same Indo-European deictics prior to the arrival of the Slavs. Whether or not it was in place prior to the arrival of Latin is currently a subject of debate, as is the question of whether the preposed manifestations of the article are earlier or later.
(c) Slavic arrived in the Balkans with postposed definiteness marking on the adjective (of pronominal, not deictic, origin), and by the time of the earliest documents, the deictics that would develop into definite articles were already undergoing (or had undergone) encliticization.
(d) Latin arrived in the Balkans with no definite articles but with postpositioning of demonstratives as an acceptable variant to a preposed demonstrative. By the time Romanian appears, it has an established postposed definite article based on the same demonstrative that supplied the preposed definite article for most (but not all) of non-Balkan Romance.
The ultimate source of this Balkan innovation has always been controversial, but the final piece of the puzzle posed by (a)–(d), and especially (b)–(d), is the analysis Hamp 1982 has offered (see §1.2.1 and §6.1.2.2.1.2) for the place name Drobeta, with its postposed definitizing element (*tā) from a substrate language. That is, since we presume that Balkan Latin could postpose deictic determiners and we know that Slavic could too, it must be supposed that something akin to what Mufwene 2001a, 2008 would call “feature selection” from among available variants (see §3.2.1.4) occurred in the ancient Balkans, with the postpositive variants being selected for in Balkan Slavic and Balkan Romance. As to the question of what drove the feature selection, Hamp’s Drobeta account provides the answer. The postposed definite article would thus have been a substratum feature that either made its way into Albanoid from this substrate as the result of language shift or was already present in Albanoid, and presumably made its way into Balkan Romance as a substrate effect.Footnote 139 Somewhat later on, Slavic speakers in contact with Balkan Romance and with Albanian would have been moved in the direction of selecting for and fixing grammatically (“grammaticalizing”) their postpositive definite article variant.
Thus the ultimate scenario for the emergence of the Balkan postpositive definite article requires a mix of language-external factors – in particular a substratum with the feature in question and reinforcement of the selection of a variant through contact – and language-internal factors, in the form of the presence of postposed definiteness marking in Slavic, and Latin, from the start. These developments thus involve a case of substratum transfer in language shift, as Thomason & Kaufman 1988 would have it, or “imposition,” in van Coetsem’s influential schema for contact-induced change (see §3.2.1.2 on these concepts). The particular transfer here would also be in accordance with the claim of Thomason & Kaufman 1988 that gross features of syntax, such as word order, carry over in substratum shift contexts (see also Joseph 1999a). One important detail to note is that native material is used for this postpositive definite article in each of the languages, so that it has the character as well of a syntactic calque (“copying,” in Johanson’s 1992 terminology; cf. also Gołąb 1976: 304 – see §3.2.1.2), and the transparent connection between the article and related demonstrative elements so evident in those Balkan Slavic varieties with multiple deictic distinctions with articles (cf. §6.1.2.1 on Macedonian t-/n-/v- deixis in its article system) makes it clear that there was at least elaboration of this feature within the individual languages. Still, whatever the contact-source of the enclitic article, it seems not to have involved the borrowing of an element’s form from one language together with that element’s prosodic characteristics, as was the case with the Balkan discourse marker de from Turkish (see §4.3.4.1 and §5.5.3), because in each language, native material is used. That is, the contact languages involved in each case (substratum starting point, Albanian and Romance influence on Slavic) provided structural templates for the diffusion of the postpositive definite marking. If it were a case like de (assuming de is indeed from Turkish), it would be surprising to see the native material at all.
Thus, the postpositive definite article can be judged a Balkanism. It is shared by a number of languages in the region and cannot be an inheritance from Proto-Indo-European. Moreover, language contact was certainly involved, even if some language-internal factors played a role. The fact that this feature may have developed in languages outside the Balkans does not vitiate the contact-based account given here. That Scandinavian languages and Armenian developed postposed definite articles (with different properties from the Balkan ones) does not change the fact that both Balkan Romance and Balkan Slavic developed postposed articles out of native materials and drawing on native syntactic patterns at precisely the time that speakers were in contact with one another as well as with the language that became Albanian. Thus Balkan Romance and Balkan Slavic can be described as convergent, and the Scandinavian and Armenian developments must be taken as an independent though parallel (even if the former may well have something to do with a similar construction in North Russian).
6.1.2.2.1.6 More on Articles: Turkish, Romani, Celtic, Romanian, and More on Macedonian
The discussion in the preceding several sections has focused on the question of postpositive definite marking, its origins, and parallels among the different languages. Despite the well-deserved attention to the development of this feature, there is more to say about articles and definiteness in the Balkans, and in this section we offer a selection of other articular developments relevant to Balkan language contact.
First, Turkish has been omitted from earlier discussion because it does not have a definite article per se, although it does have a form of definiteness marking. In Turkish (and Turkic) grammar there is a suffixal case marker for definite or specific direct objects, a feature attested in the oldest documents (see Erdal 2004: 366, cited above in Footnote footnote 26). This sort of definiteness marking does not appear to have interacted with definiteness in the other Balkan languages nor to have been affected by those other languages. Moreover, West Rumelian Turkish use of deictics does not appear to differ significantly from the rest of Turkish. In general, Turkish uses deictics in some environments where languages with morphologized definite articles use articles.
Second, Romani appears to have developed a definite article in contact with Byzantine Greek out of the same demonstrative material that served as the basis in the other Indo-European languages. Some Romani articles look like borrowings from Greek, e.g., m.nom.sg o / f.nom.sg i, but the oblique forms such as le and la in Vlax dialects demonstrate that the Romani articles are derived from native demonstratives, reflecting the regular change of *t > l, which occurred prior to contact with Greek (Sampson 1926: 151–153, 247; Boretzky 2000a; Matras 2002: 96–98). The formal resemblances to Greek are therefore coincidental, but it was contact with Greek that probably triggered the transformation of native material into definite articles, and Romani usage patterns very much like Greek (Boretzky 2000a; Matras 2002: 96–98).Footnote 140 A subsequent development is the tendency of some Romani dialects outside the Balkans in contact with languages lacking definite articles to lose them, as happened in Northeastern Romani dialects in contact with Baltic and North Slavic (Tirard 2019: 438). See §6.1.2.3 on multiple definiteness marking.
Third, there are some intriguing facts regarding Celtic in the Balkans. These are presented earlier in §1.2.1.8, but they have added curiosity value here in the light of the discussion about the prehistory of the Balkan definite article. In particular, the Grafenstein inscription has the sequence of letters ollo=so, which some scholars believe contains a Celtic word for ‘all’ (cf. Old Irish oll ‘ample’) followed by a postpositive deictic (cf. Sanskrit sa). As emphasized in §1.2.18, so much about this inscription is unknown that it cannot be used to further our understanding of definiteness in the Balkan languages, but this interpretation, if correct, affords room for speculation as to whether the “Drobeta substrate language” was Celtic, whether the Postclassical Greek development with the article following ‘all’ is at all connected with the pattern seen here, and so on.
There are also three items from Macedonia worthy of mention. One is that in light of the Agía Varvára facts cited above, it can be noted that Koneski 1981: 128 mentions that doubled articles occur in southern Macedonian under Greek influence. Thus this pattern is borrowable/transferable. On the other hand, expressive doubled articles can occur in Macedonian in the context of double determination, e.g., ovie našive polupismenive (lit., ‘these our.def semiliterates.def’) ‘these semi-literates of ours’ (Prizma 2015 as cited in Friedman 2019d), although this usage is considered marginal.Footnote 141 In Golobrdo (Alb Golloborda) and Gora, however, doubled articles can occur, e.g., našite starcite ‘our old folks’ (Steinke & Ylli 2008: 74), starete zapišanete lafoji ‘the old [the] written-down words’ (Nomachi 2019: 193). Moreover, there is an example from a Bulgarian corpus of spoken dialogue (Разговорна реч [bgspeech.net 2009–2024] 28.txt): tvoite dvete golemite ‘your two big [ones; plastic boxes].’ See Footnote footnote 120 for an example from Torlak BCMS. Early Modern Romanian also supplies such examples, e.g., păntru sufletul răposatului jupânului Predei ‘for the soul of the late master Preda (lit., ‘for soul.acc.def late.gen.def master.gen.def Preda.gen.def’) (Nicolae 2015: 25).
A second item involves the borrowing of actual material that marks definiteness, rather than the templates for structural borrowing discussed so far. Capidan 1940: 92, note 1 mentions that in Meglenoromanian a postposed definite article based in form on the Slavic marker is to be found, as in seara-ta ‘the evening,’ where the -ta seems to be a wholesale borrowing from the Macedonian feminine singular definite article, -ta.
Finally, in the area around Thessaloniki (e.g., Liti (Mac Ajvatovo); Adamou 2006), the local Macedonian speakers, who are now bilingual in Greek (and have been most likely for only a few generations), refer to their language as našta, syncopated from Slavic naša ‘ours’ plus the feminine singular article -ta, feminine so as to agree with an understood Greek feminine head noun γλώσσα ‘language’; this form has come to be treated, however, as unanalyzable, and is given a definite form naštata (e.g., znaš naštata ‘You know Macedonian’) with another feminine article suffixed to it.Footnote 142 This sort of reanalysis, while not at all unusual cross-linguistically,Footnote 143 is interesting in the context of the origin of the postpositive definite article in the Balkans because it is just like the Drobeta reanalysis. With Drobeta, a definite article in one language was interpreted as part of a stem, and the same has happened with našta-ta. This example, then, gives further plausibility to Hamp’s account.
6.1.2.2.2 Preposed Articles – An Ideological Interlude
While the postposed definite article of the Balkans struck nineteenth-century German- and Romance-speaking scholars as somehow “unnatural” and remarkable and thus requiring explanation, the emergence of preposed definiteness in the region has received little attention beyond its typological (i.e., achronic) occurrence. That Ancient Greek developed such an article is historically clear (see §6.1.2.2.1.4). Whether the development had anything to do with contact with non-Greek languages is obscured by lack of plausible documentation.
In the case of Albanian, a major question, as discussed in §6.1.2.2.1.2, is whether its preposed determiner-like (thus, article-like) element (the particle of concord – see §6.1.2.2.1.2) developed first and subsequently became postposed or vice versa. The fact that the development of a preposed article is clearly traceable within Ancient Greek, and that likewise the postposed article is clearly a late development in Balkan Latin, combined with the fact that postposed definiteness marking is clearly reconstructible for Common (Balto-)Slavic (see §6.1.2.2.1.1 – the current BSl definite article per se is later), has meant that a certain amount of language-ideological baggage attaches itself to speculations about Albanian. A postposed definite article could, on the one hand, be attributed to a Slavic tendency (Mladenova 2007, but cf. Sobolev 2009b contra). On the other hand, it has commonly been attributed to a pre-Latin language of the Balkans north of Hellenic, i.e., Illyrian (given the complete lack of inscriptions we use this term as a convenient regional label) or Thracian or Daco-Moesian (Kopitar 1829, S. Mladenov 1929, and others) that is unattested or inadequately attested, unless Hamp’s 1982 reconstruction of the toponym Drobeta as *druṷā-tā is accepted (see §1.2.1 and §6.1.2.2.1.2).
From a language-ideological point of view, if Albanian is descended from or a cousin of Illyrian, which latter, everyone agrees (Woodard 2004: 7–9), was different from Thracian (sensu lato), and if the preposed particle of concord of Albanian is ancient, then a preposed definite article could be an Illyrian-Greek innovation (or even an Illyrian-to-Greek contact phenomenon, with Homeric usage, like Czech or Slovene vis-à-vis German today, showing the beginnings of articular use of an earlier demonstrative element). The Hellenic dark ages – i.e., the period between the disappearance of literacy from contact with Aegean languages (c. 1200 BCE) until the reappearance of literacy in contact with Semitic (c. 800 BCE) – allow for considerable speculation. Here it is to be remembered that on both occasions, Greek literacy was the result of the contact of a pre-literate language (Greek) with literate cultures (Aegean and Semitic). Given the preposed definite article of Northwest Semitic, which is also of deictic origin (Sáenz-Badillos 1993: 58), on the one hand, and the Semitic origin of the Greek alphabet, on the other, it is easily argued that Greek’s preposed definite article was influenced by the already established usage of the Semitic languages with which Greek was in contact.
Owing to the fact that linguistic convergence often reflects unequal power relationships, combined with the fact that power is often associated with prestige, questions of the “origin” of this or that Balkan linguistic phenomenon become ideologically freighted. And since the postposed definite article was one of the first things that struck Western European observers (Leake 1814; Kopitar 1829) as linguistically unusual about the Balkans, its origin is especially cathected. For Romance, it is a clear innovation in Balkan Romance and thus has been connected to Dacian (though see §6.1.2.2.1.3 for in-depth consideration). For Balkan Slavic, the tendency towards postposition is old, and thus the arguments can go either way (though see §6.1.2.2.1.1 for full discussion). For Greek, preposing is obviously old, and as for Judezmo, it arrived in the Balkans with the Western Romance preposed article in place (see §6.1.2.2.1.3). The Romani preposed article arguably developed during its contact with Byzantine Greek, possibly outside of the Balkans, but used native material (see §6.1.2.2.1.6). Thus, Albanian is alone among the Balkan languages for which the articles are in place, but earlier attestations are lacking, so that an ideological appropriation of the preposing-in-origin versus postposing-in-origin issue is, in a sense, not surprising.
From this it emerges that preposed articular manifestations are not marginal phenomena in the Balkans, and for Romani they are and for Greek they might be, involved with language contact. Nonetheless, it is clear that the postposed articles have commanded more attention over the years, and perhaps rightly so, for that is where more data are available and more complex questions can be asked and considered.
6.1.2.3 Double Determination
The term double determination represents two distinct but related notions with respect to noun phrases. The first has to do with the presence of a definite article on a noun modified by a demonstrative pronoun or adjective, while the second refers to the multiple presence of definite articles or definiteness markings with more than one element in a noun phrase. Double determination itself is arguably a Balkanism, although the demonstrative systems of the Balkan languages show considerable variation.
Double determination with demonstratives occurs in Greek, Balkan Romance, Balkan Slavic, Albanian, and Romani, although the rules as well as the relative frequency and register of the construction vary. In Greek it is obligatory, so that for ‘this man,’ one finds only αὐτὸς ὁ ἄνθρωπος (lit., ‘this the man’) or ὁ ἄνθρωπος αὐτός (‘the man this’) but never *αὐτὸς ἄνθρωπος (‘this man’). In Romanian, the article is not used if the demonstrative is preposed, but it must be used if it is postposed (with the demonstrative taking the so-called deictic particle -a): om-ul acesta (lit., ‘man-def this’,) but acest om ‘this man,’ cf. Aromanian aistã carte ‘this book’ and cartea aistã ‘book.def this.’ Megenoromanian frequently has double determination, e.g., tsista lup-u ‘this wolf-def,’ but indefinite nouns also occur with demonstratives, e.g., tsista drăc ‘this devil.indf.’ In both Greek and Balkan Romance (except perhaps in Meglenoromanian, which in this respect probably reflects Macedonian influence), the rules governing demonstrative + definite are grammatical rather than pragmatic.
In Albanian and Balkan Slavic, double determination is facultative, i.e., pragmatically determined. In general, it conveys emphasis or subjectivity. However, there are some apparent grammatical restrictions on its occurrence that form a continuum from Albanian – with the fewest restrictions – across Macedonian (including Goran, see Nomachi 2019) to Bulgarian, which seems to have the most. In Albanian, double determination can occur with both substantives and adjectives in a single NP, e.g., ky djal ‘this boy,’ ky djal-i ‘this boy-def,’ ky djal-i im i zgjuar ± i ‘this clever boy of mine’ (lit., ‘this boy-def my pc clever ± def’).Footnote 144 In Macedonian, it can occur with either substantives or adjectives, but, normally, not both together, e.g., ovie idioti-ve ‘these idiots’ or ovoj naši-ov čovek ‘this guy of ours’ (but cf. also ovie naši-ve polupismeni-ve ‘these semi-literates of ours’ cited in §6.1.2.2.1.6; all examples from Prizma 2015, Friedman 2019d). In Bulgarian, such double determination is only acceptable to most speakers with an adjective, e.g., tazi nejna-ta banica e strahotna ‘that her-def banitsa [a kind of pastry] is awesome,’ but not *tozi idiot-ăt ‘this idiot’ (Rudin 2019; Friedman & Rudin 2021). Torlak BCMS has taj postara-ta ‘that the.older.one-f.def’ (A. Belić 1905: 447), za onúo košárku-tu ‘for that the.basket.acc,’ tie pesme-te ‘these the.songs,’ toa vaši-ja dom ‘this your house’ (Sobolev 1994: 185–186).Footnote 145
Many Romani dialects in the Balkans permit but do not require the use of a definite article with a preposed demonstrative, in which case the article must precede the substantive, but if the demonstrative follows, the substantive must be definite in some dialects, e.g., ‘this person,’ one finds kadava [o] manuš ‘this [the] person,’ and, o manuš kadava ‘the person this’ (Igla 1996: 165; cf. also Boretzky 1993, 1994; Cech & Heinschink 1999: 34, 38; Sechidou 2011: 62; Cech et al. 2009; Tirard 2019). Double determination or the order noun-determiner is pragmatically more thematic in discourse.Footnote 146 However, a postposed demonstrative without a definite article is possible in at least some dialects. As argued by Boretzky 1994: 55; Igla 1996: 165; Matras 2002: 97, 2004: 78–79; and Tirard 2019: 125, the construction of the type dem + noun is inherited, and both dem + def + noun and noun + def + dem are borrowed. And, as Tirard 2019 makes clear, the construction is found throughout the Vlax and Balkan Romani dialects of the Balkans, but not beyond the Balkans (except in the case of recent migrations).
As for the history of this construction within the Balkans, all that can be known for certain is that the double determination pattern was a regular part of Classical Greek nominal morphosyntax, but was not present in Latin or early Indic, languages in which there was no definite article. In Common Slavic, as represented by OCS and as noted above, the inherited pronoun *jĭ was suffixed to adjectives as a definite marker. In the presence of a deictic pronoun, it was normal for the adjective to be definite (Flier 1974: 158–160). The survival of the reflexes of these definite adjectives, although their meanings and forms have changed in various ways in various languages, guarantees that this development occurred prior to the Slavic migration to the Balkans. Still, the current situation in Balkan Slavic is considerably later, and it arose specifically in the Balkans. With regard to Albanian, the pragmatic conditioning looks like Romani and Balkan Slavic, but, as so often happens, the lack of time depth leaves things uncertain. Tirard 2019 has argued that Greek is the ultimate source of the construction in Romani, and while other contact languages may also have had input, given that the Romani definite article was calqued on Greek (see §6.1.2.2.1.6), that language was certainly at least a source. On the other hand, the differences in detail among the languages that have this kind of double determination might argue in favor of independent origin for some of them. And certainly from a typological perspective, the sort of redundancy that double determination instantiates is not surprising; intra-NP redundant expression is found in the use of plural noun forms with numbers in most Indo-European languages as a similar phenomenon, and, even more to the point, in further specifications with demonstratives, such as English this here (horse, man, etc.), or French celle-ci ‘this here.’Footnote 147 These are all suggestive of a universality to double determination (and multiple deixis) that would challenge its status as a possible Balkanism. Still, the Romani situation vis-à-vis Greek (and Romanian, see Tirard 2019 and Adamou & Matras 2021) is clear. Moreover, the continuum from Albanian through Macedonian to Bulgarian (especially if the example ovie našive polupismenive ‘these illiterates of ours’ cited in §6.1.2.2.1.6 is more typical of western Macedonian) is highly suggestive of contact effects.
Double determination of a somewhat different sort occurs in eastern Torlak BCMS in feminine substantives ending in a consonant. In such cases the article is sometimes doubled, suggesting that the first occurrence was interpreted as constituting an oblique stem, e.g., nočtútu ‘the night,’ večertútu ‘the evening’ (A. Belić 1905: 447). If that is indeed the correct interpretation of these forms, then they can be added to the instances of reanalysis noted in §6.1.2.2.1.6 in connection with the term našta in Macedonian dialects in Greece (Adamou 2006). Such reanalyses are not uncommon cross-linguistically and actually may best be considered double determination only in an etymological sense; if an element that was once a definite marker is incorporated into a stem, it thereby loses its semantic definiteness. In some instances, though, a remnant of a definiteness marker may retain at least an association with definiteness even if it is not an overt maker per se. This seems to be the case with remnants of the Common Slavic “long adjectival” forms in some of South Slavic (see §6.1.2.2.1.1), where, for instance in Macedonian and Bulgarian, the main way in which the long form of an adjective survives is when it is preposed and occurs with a definite article, e.g., Mac golemiot/Blg golemijat brat ‘the big brother’; in such a case, inasmuch as the independent use of the long form (here, golemi) is quite restricted, (see §6.1.2.2.1.1), one could argue that the etymological definite marking is associated synchronically with definiteness without necessarily directly marking definiteness itself.
Finally, mention must be made here of the Greek polydefinite construction and Balkan-internal parallels to it, discussed in greater detail in §6.1.2.2.1.4. In this construction, there is a repeated article in noun phrases with post-nominal adjectival modification, as in (6.66):
(6.66)
το σπίτι το μεγάλο the.sg house.sg the.sg big.sg ‘the big house’ (lit., ‘the house the big’)
In this form, it was calqued into Romani. Today it occurs in both Balkan and Vlax Romani dialects in the Balkans, some of which have expanded on this construction and now have syntagms that cannot occur in Greek. Taking the lexical items in (6.66), the literal Romani translation, o kher o baro, works just as well in various dialects of the Balkans. One could also have o baro kher = το μεγάλο σπίτι. However, a number of Romani dialects (Greece, Albania, North Macedonia, etc.) also have o kher baro = *το σπίτι μεγάλο (cf. Tirard 2019: 492; Cech et al. 2009: 230; and also the references therein). It is thus the case that, having begun with the basic order def adj noun, Romani calqued def noun def adj, and also ended up with def noun adj, which may or may not have been a subsequent development, but is especially prevalent where current contact languages tend to favor this last order (Albanian, Romanian; see Adamou & Matras 2021 for Romani in Romania), although as a marked order it occurs in dialects where the order is also marked in the main contact language(s), e.g., in North Macedonia. See also the references earlier in this section and the discussion in §7.4.1.1.
The Modern Greek construction continues one occurring in Classical Greek, so the directionality of the spread in the case of Romani is clear. Moreover, as noted above, Koneski attributes multiple definite articles in dialectal Macedonian to Greek influence. There are also parallels elsewhere in the Balkans, in the form of multiple marking of definiteness in a noun phrase without necessarily involving the Greek-style repetition of the article per se. In particular, the so-called adjectival articles of Albanian and Balkan Romance can be interpreted as instances of added marking of definiteness, at least from an etymological standpoint; see §6.1.2.2.1.2 and §6.1.2.2.1.3 respectively for more discussion.Footnote 148
Despite the likely universality for the functional value of redundancy and expressiveness that can lead to multiple indications of definiteness and deixis in a word or phrase, some of the forms of double determination in the Balkans taken together would seem to be a good candidate for consideration as a Balkanism, even if one of limited scope. In particular, the double determination involving a demonstrative and an article shows a geographical distribution and a comparative aspect indicative of a specifically contact-related phenomenon. Sandfeld 1930: 122 notes the convergence of double determination (dem + def) in Macedonian, Albanian, and Greek. As discussed above, the greater restriction in Bulgarian is consistent with an areal spread from the southwest convergence area. Moreover, despite Sandfeld’s 1930: 3 discounting of Romani, it is very clearly part of the areal picture. To this we can add that Turkish, too, has the possibility of dem + def insofar as the demonstrative (dem) can cooccur with the definite accusative, e.g., al bu elmayı! ‘take this [specific] apple!’
Thus, the particular combination of demonstrative and definite article can be judged a Balkanism. Taken together with the clear cases of diffusion of specific patterns, in particular Greek articular polydefiniteness, it shows the importance of double determination in the Balkans. From a functional perspective on language contact, this result also makes sense, since double determination is a type of redundancy, and redundancy, in contact situations, makes for clearer and more effective, i.e., more successful, communication.
6.1.2.4 Indefiniteness
The morphosyntactic realization of indefiniteness in the Balkans, as suggested in §6.1.2.2, offers an important corollary to the realization of definiteness. The most significant aspect of Balkan indefiniteness has to do with the numeral ‘one,’ though there are some other lexical facts of relevance too.
6.1.2.4.1 ‘One’
The use of an unstressed form of the numeral ‘one’ as an indefinite article is characteristic of all the Balkan languages. Grammatical accounts of Albanian (Newmark et al. 1982: 150–151), Greek (Householder et al. 1964: 96), Turkish (Lewis 1967: 53–54), and Balkan Romance (Graur et al. 1966: 108–111 and Atanasov 1990: 201) explicitly describe etymological but unstressed ‘one’ as an indefinite article in their respective languages, and while Aromanian grammars do not overtly label un[u]/[u]nă ‘one.m/f’ in this way, Vrabie 2000: 101 glosses it with English ‘an.’ Romani uses jekh ‘one’ in this same fashion (Friedman 2001d). Balkan Slavic shows this usage too, but presents a picture that is complicated by puristic and prescriptive pressures. The various languages do not behave identically in the details of their use of this form of ‘one,’ but in the modern languages, usage in Turkish, Albanian, and Balkan Romance is at a similar level of frequency to that of English, whereas text-count frequency shows usage in Balkan Slavic and in Greek to be approximately half that of the other Balkan languages; usage in Balkan Romani patterns with Balkan Slavic and Greek (Friedman 2003c treats all the languages).
Such a development is typologically common and found in many languages, e.g., in Germanic (cf. English a(n), German ein-), non-Balkan Romance (cf. French un/une, Italian uno/una), and Basque, where bat, means ‘one’ as an independent form but serves as an indefinite article when postposed (cf. Sinhalese as described in §6.1.2.4.1.4, Footnote footnote 152). Nonetheless, the usage in the Balkan languages is arguably a Balkanism given the timing of its emergence and its distribution. The facts from the various languages are surveyed here, followed by a summary of the history and status of this feature.
6.1.2.4.1.1 Albanian
For Albanian unstressed një (Geg nji) occurs as indefinite article in the earliest documents, where it declines (the oblique form being njëj/njaj/njij before the gen/dat/abl; Sh. Demiraj 1985: 303–304).Footnote 149 In the modern language, një does not decline when used as an indefinite article. When një is used with certain locative prepositions it can specify an indefinite reading, e.g., Hipi mbi çati ‘He climbed on the roof’ vs. Hipi mbi një çati ‘He climbed on a roof’ (see also §7.9.2). The indefinite article is used with predicate nominatives only when they are modified: Ai ishte djalë ‘He was a boy’ vs. Ai ishte një djalë i vërtetë ‘He was a real boy’ (Newmark et al. 1982: 150–151).
6.1.2.4.1.2 Balkan Romance
‘One’ was not used as an indefinite article in Classical Latin, although such usage does begin to appear in Vulgar Latin by the fourth century (Mihăescu 1978: 616), and it is a common heritage of the Romance languages. Grammars of Balkan Romance languages describe etymological but unstressed ‘one’ as indefinite articles, e.g., Graur et al. 1966: 108–111; Stan 2013b: 291–294; Farkas 2013, for Romanian; and Atanasov 1990: 201 for Meglenoromanian. As noted above, Aromanian grammars do not treat the status of un[u]/[u]nă ‘one.m/f’ as an indefinite article explicitly, although Vrabie 2000: 101 glosses it with the English indefinite article and it is clear from publications and texts that it has this function.
In Romanian, ‘one’ is used with predicate nominatives to distinguish referential from generic/attributive, similar to the usage in Albanian cited above, e.g., Ionescu este clovn ‘Ionescu is a clown [by profession],’ Ionescu este un clovn ‘Ionescu is a clown [a prankster]’ (Graur et al. 1966: 109).
6.1.2.4.1.3 Greek
The use of ‘one’ as an indefinite marker is unattested in Ancient and New Testament Greek.Footnote 150 It first appears as such in Byzantine Greek (cf. Horrocks 2010: 330–331), and grammars of Modern Greek describe etymological but unstressed ‘one’ as indefinite articles (Householder et al. 1964: 96; Holton et al. 1997: 282–285). Householder et al. 1964: 96 write (Greek given as in original, transliterated as if Ancient Greek):
The indefinite article is used more sparingly in Greek than in English; it is not used for example, with predicate nouns, often not with indefinite direct objects, and generally not in proverbs and popular sayings when an entire class is meant and not a specific member of that class: eîmai foitētè:s ‘I am a student’; khtízoune spíti ‘They are building a house’; katharòs ouranòs astrapès dè phobâtai ‘A clear sky is not afraid of lightning’. Examples: sâs ze:tâ énas kúrios kaì mià kuría ‘A gentleman and a lady are asking for you’; miâs fíle:s mas tê:s éklepsan té:n tsánta ‘They stole the handbag of a friend of ours’; vré:kame éna skúlo stò drómo kaì tòn phérame spíti ‘We found a dog on the street and brought him home.’; autò tò kapélo eînai enòs kuríou poù ménei edô ‘This hat belongs to a gentleman who lives here.’
According to Kazazis (p.c.), modified predicate nominatives ordinarily do not take ‘one,’ as in είμαι καλός φοιτητής ‘I am a good student’ or είμαι φοιτητής της ανθρωπολογίας ‘I am a student of anthropology.’ On the other hand, as an identifying referential, ‘one’ occurs in contexts such as the following: είμαι ένας φοιτητής της ανθρωπολογίας στο Πανεπιστήμιο Θεσσαλονίκης και σας γράφω για να σας ζητήσω μια συμβουλή ‘I am a student of anthropology at the University of Thessaloniki and I am writing to you to ask for your advice.’ Similarly, to the question Ποιος είναι? ‘Who is it?’ (e.g., on the phone, at the door) one can answer [είναι] ένας φοιτητής ‘[It’s] a student’ (cf. Holton et al. 1997: 283). It is worth noting in the context of the indefinite article in Balkan Slavic (see §6.1.2.4.1.6) that the use of ‘one’ as an indefinite marker was discouraged for Katharevousa (Kuhlmann 1997: 92) and also in the Modern Greek translation of the New Testament (Friedman 2003c).
6.1.2.4.1.4 Romani
Sanskrit did not use ‘one’ as an indefinite article although it has emerged as such in some modern Indic languages (Masica 1991: 248–250). Romani usage is an independent development and varies according to contact languages; those Romani dialects in contact with languages that have an indefinite article are likely to use jekh ‘one’ in the same way.Footnote 151 For the Balkans, indefinite article usage is attested in the oldest Balkan Romani text, from the seventeenth century (Friedman & Dankoff 1991), and current usage resembles that in Greek or Balkan Slavic (Friedman 2003c).Footnote 152 Boretzky 1994: 58 reports that in the Kalderash (Vlax) Romani of Vojvodina ek/ək/k – reduced forms of jek ‘one’ – are used as indefinite articles, e.g., De ma jek phabaj vs. De ma k phabaj ‘Give me one apple’ vs. ‘Give me an apple.’
6.1.2.4.1.5 Turkish
The use of ‘one’ as an indefinite article is attested in the oldest Turkic documents (eighth century CE and onward; Tekin 1968: 145; Erdal 2004: 359). Grammars of Modern Turkish describe unstressed bir ‘one’ as an indefinite article, e.g., Lewis 1967: 53–54, 248; Göksel & Kerslake 2005: 164–166, 202f., 209, 372f., 381. The position of bir (colloquial bi when functioning as an indefinite article) within the NP determines the nature of its specification vis-à-vis a modifier, e.g., güzel bir bahçe ‘a beautiful garden’ (as opposed to an ugly one), bir güzel bahçe (as opposed to a beautiful meadow or ugly forest). This kind of usage can occur in other Balkan languages, but only colloquially.
6.1.2.4.1.6 Balkan Slavic
The numeral ‘one’ (Mac éden, édna, édno, Blg edín, edná, ednó m, f, n, respectively, with the potential for reduced forms) can function as an indefinite article in much the same way and with approximately the same frequency as the Greek equivalents (Friedman 2003c). The more referential the function, the more likely the indefinite article is to occur. From a Balkanological standpoint, it is clear from the documentation in Old Church Slavonic, as well as various other Slavic languages, that the Balkan Slavic usage arose during the period and in the context of Balkan linguistic contact. As with the use of etymological ‘want’ to mark futurity or the replacement of infinitives with dms-clauses, the use of ‘one’ as an indefiniteness marker extends into BCMS territory. Moreover, the analysis of the uses of ‘one’ as an indefinite marker in Balkan Slavic is a site of normative and puristic anxieties and therefore requires some additional explanation.
As indicated in the preceding paragraph, the use of ‘one’ as an indefinite marker was not inherited from Common Slavic, and, moreover, it did not develop as such outside the Balkans.Footnote 153 For Macedonian, eden has been analyzed as an indefinite article by Topolińska [Topolinjska] 1981/82, Weiss 1996, Friedman 1993d: 268, 291, and Minova-Gjurkova 1994: 59–60, 118–129.Footnote 154 Topolińska [Topolinjska] 1981/82: 712 makes the point that Macedonian eden is a semantically bleached referential and generic marker whose use is broadening among younger Macedonians (cf. also Naylor 1981/82: 538).Footnote 155 Moreover, Friedman 1993d: 291 makes the point that, as in Greek and Albanian (Kazazis & Pentheroudakis 1976), Macedonian eden in its specific-referential function, is integrated into the grammar to the point that it can even trigger object reduplication, as in example (6.67), cf. Koneski (1967: 231–222):
(6.67)
Ja baram edna marka, no ne ja najdov. I seek one stamp but neg it found ‘I’m looking for a [specific] stamp, but I couldn’t find it.’
Nonetheless, the use of ‘one’ as an indefinite marker in Macedonian (as in the other Balkan Slavic languages, see below) is considered colloquial and is often edited out of publications by copyeditors.
For Bulgarian, Friedman 1976b summarizes the debate up to that time and reaches the conclusion that ‘one’ is indeed an indefinite article in that language. Mayer 1988: 121 reaches the same conclusion, noting that “the use of edin is obligatory in indefinite NPs expressing specificity when the NP does not carry logical stress.” Nonetheless, Avgustinova 1998: 15 writes: “The existence of an indefinite article in Bulgarian, addressed, for example, in Friedman 1976b, is still a controversial issue and a matter of on-going linguistic discussion,” and Bojadžiev et al. 1998: 470 state: “ … it is not logical to accept the existence of an indefinite article of the type edna kniga [‘a/one book’] … ”Footnote 156 Here Brezinski’s 1969: 49 normative criticism of the “unmotivated” use of edin in Bulgarian, claiming that the usage is the result of influence from the French and German indefinite articles, reveals the underlying puristic anxieties associated with this debate. Russian does not have this use of ‘one’ (odin, odna, odno, cf. Nicolova 2000), and the usage is therefore perceived as “un-Slavic.” We can say, however, that the influence is not from French or German but rather from the surrounding Balkan languages, especially in view of the fact that the usage is typically colloquial and that children are taught in school to avoid it.Footnote 157
In the former Serbo-Croatian, the use of jedan ‘one’ and its co-forms (jedna, etc.) as a marker of indefiniteness was more characteristic of the Serbian standard than the Croatian, or perhaps was more the object of prescription in Croatian but description in Serbian. Maretić 1963: 510 writes for Croato-Serbian: “ … many modern writers spoil the language by using the numeral jedan without any need on the model of German ein, French un, Italian uno.”Footnote 158 This contrasts with Stevanović 1986: 313 who writes for Serbo-Croatian: “The numeral jedan is used very often in our language not as a numeral but more as a kind of indefinite article.”Footnote 159 Hinrichs & Hinrichs 1995: 55–57 write that indefinite use of jedan is more common in Serbian varieties of BCMS, but see now also Belaj & Matovac 2015 on Croatian and elsewhere.
6.1.2.4.1.7 The Indefinite Article as a Balkan Feature
As can be seen from the individual language descriptions, the use of ‘one’ as an indefinite article was not present in Sanskrit, Ancient and New Testament Greek, Classical Latin, and Old Church Slavonic. It is attested in Old Turkic, Byzantine Greek, and later Vulgar Latin. Moreover, in terms of frequency, contexts for Romance usage, both Balkan and non-Balkan, are about twice as frequent as in Greek (Friedman 2003d). For Albanian and Romani, our earliest texts are early modern, and so there is no way of knowing how old the phenomenon is, but it is striking that Albanian patterns with Romance (and Turkic) and Romani more with Greek. Balkan Slavic, too, patterns with Greek, and indefinite article usage in that case unquestionably arose in the context of Balkan language contact. Whether Latin/Romance usage influenced Greek or Albanian is moot, but Turkish arrived in the Balkans with its indefinite article firmly established. (For a discussion of Bulgarian in comparison with Russian see Gornisheva 2016.)
6.1.2.4.2 Other Indefinites
Besides the more grammatical realizations of indefiniteness in the form of an article, there are several realizations to note that are more lexical in nature, and are thus treated in Chapter 4; see especially §4.3.3.1.2 on the fate of indefinite pronouns in the Balkans, with brief additional remarks in §4.3.3.1.1.
6.1.2.5 Morphosyntactic Topicality and Focus
Topicality and focus are marked by various means in the Balkan languages, including syntactic, prosodic, lexical, and morphosyntactic. With regard to syntax, the marking involves shifts in word order (e.g., in Greek, sentence-initial constituents typically being topical, while in Romani, object-verb order signals topicality for the object), as argued by Arvaniti & Adamou 2011; see also the brief discussion in §7.4.2); with regard to prosody, intonation can be used, as in Romani but so also stress shift within a word for focus, deaccenting for topicality (see Adamou & Arvaniti 2010; Arvaniti & Adamou 2011); and, with regard to lexical markers, in Slavic, focus can be marked in yes–no questions by the enclitic word li, which, as shown in §4.3.3.5 and §6.2.5.3, has been borrowed into Romani, and Romani has also borrowed Turkish da (Adamou & Arvaniti 2010; Arvaniti & Adamou 2011; see also §4.3.4.1.2). Turkish can also use its interrogative marker mI as a focus marker, a process already attested in Old Turkic (Erdal 2004: 412).
Of particular relevance here are two morphosyntactic processes that are employed in topic and focus marking. One, object reduplication, is a construction in which a weak object pronoun doubles and cross-references a direct or indirect object; although it can certainly be considered morphosyntactic in nature, inasmuch as it involves weak object pronouns (clitics), it is given a full treatment under the rubric of syntax, being discussed in §7.5.1. The other is the process discussed above involving the marking of a direct object by an element that was originally a preposition. As noted in §6.1.1.1.2, this marking found in Balkan Romance and to a limited extent in Balkan Slavic, can be considered a type of “differential object marking.” Interestingly, Hill 2013: 145 argues that for Romanian, the function of object marking with pe in Early Modern Romanian was “to achieve a Contrastive Topic reading on the direct object,” but that now, pe has undergone a category change from preposition to functional morpheme and accordingly serves as “a backgrounding device for the [+human] noun in the discourse.” Though a somewhat restricted process in the Balkans, nonetheless this deployment of the morphosyntactic marking of direct objects shows how speakers make use of means at hand to indicate pragmatically important distinctions.
6.1.3 Gender
The intersection of nominal gender and language contact in the Balkans does not yield a particularly rich set of developments. Still, there are some that are worthy of discussion even if the role of language contact is not always clear. We address here the question of gender distinctions, where contact may or may not have played a role, and a mix of minor developments where the causal role of contact is clearer. For our purposes, the term gender covers two types of agreement classes: controller gender and target gender (Corbett 1991: 147–154). Controller gender refers to the classes to which nouns need to be assigned in order to account for agreement patterns, and target gender refers to the morphology of the agreeing forms (e.g., adjectives).
6.1.3.1 Gender Distinctions
The languages differ in the extent to which and ways in which the set of earlier gender distinctions are treated.Footnote 160 Greek and Slavic, for instance, preserve the Indo-European three-gender opposition of masculine-feminine-neuter. The patterns of syncretism generally put neuter with masculine, with neuter as the unmarked gender, e.g., the gender that is used when the subject belongs to a genderless class such as a clause. Balkan Slavic, including Torlak, however, unlike the rest of BCMS as well as East Slavic and Polish (but like Slovene and the rest of West Slavic), preserves the syncretism of neuter with feminine in dva-dve ‘two [m-f/n] rather than shifting the neuter to masculine, a conservatism that is suggestive of a shared retention in these speech communities against all odds in a contact zone; Balkan Romance and Albanian both show syncretism of neuter and feminine in certain contexts, so Albanoid speakers may have contributed to this retention (though it is lexically very limited in Balkan Romance). It is worth noting that in Balkan Slavic, as in Greek and unlike, e.g., some of North Slavic, neuter gender can denote living beings, e.g., BSl momče, Grk αγόρι ‘boy’ (cf. Aronson 1964). Mladenova 2001 argues that the use of neuter gender for mature humans in Greek and Balkan Slavic is a Balkan innovation. Turkish lacks grammatical gender completely and Romani has only the masculine/feminine opposition. In the case of Romani, the assignment of neuter nouns to either masculine or feminine usually agrees with the rest of Indic and is generally assumed to have taken place during the Middle Indic period before Romani left India (Matras 2002: 72; Oslon 2012; Boretzky 2013). The masculine is the unmarked gender.
Like the rest of Romance, Balkan Romance has only two target genders: masculine and feminine. The same can also be said for attested Albanian. In both languages, however, it is the feminine gender that is unmarked (but see Corbett 1991: 213–214 for Romanian [but not Albanian]). This status is shown, for instance, by the use of the feminine weak object pronoun in the Object Reduplication construction (see §7.5.1), as exemplified in (6.68) for Romanian (from Pană Dindelegan 2013: 138), and the feminine particle of concord in (6.69) for Albanian with a predicate adjective:
(6.68)
[Că este leneş]i (astai) oi ştiu de mult that is lazy that it.f.acc know.1sg for long ‘I have known for a long time that he is lazy’ (lit., ‘I-know it (that) that he-is lazy … ’)
(6.69)
Është e qartë se … is pc.f.nom clear that … ‘It is clear that … ’ (lit., ‘is clear(F) that … ’)
Romanian is the only Romance language to use the feminine gender in this way; as Pană Dindelegan 2013: 138 puts it, “The use of the feminine clitic o with a neutral value is distinctive of Romanian.” Moreover, Greek uses the neuter here (i.e., το ξέρω ότι … ‘it.3sg.n know.prs.1sg comp … ’) and to the extent that such reduplication can occur in Slavic – with a clausal object, it is a nonstandard usage in Macedonian – only neuter would be possible (e.g., go znam deka … ‘it.n know.1sg that … ’). But for the expression of ‘the good’ as an abstraction, the neuter dobroto would be used in Macedonian, and it corresponds to the Albanian feminine e mira. It is thus tempting to see in the Albanian–Romanian convergence on this very specific point a grammatical feature that links these two languages in a possibly chronologically quite deep way (see §1.2.1.4 and §4.2.1.1 on relevant lexical evidence). Moreover, the correspondence of feminine to neuter in Aromanian/Macedonian perfects is also telling (see below and §6.2.3.2.1).
The situation with Balkan Romance and Albanian, however, is somewhat more complicated. In particular, it is necessary to posit three controller genders for them. Both Balkan Romance and Albanian have nouns that control masculine agreement in the singular and feminine agreement in the plural. In Albanian, nouns of this type are always inanimate, and in Balkan Romance they are usually, but not always, inanimate (Newmark et al. 1982: 132; Gönczöl-Davies 2008: 18). The third controller gender is often referred to as neuter in descriptions of Albanian and Balkan Romance, but, unlike the neuter of Balkan Slavic and Greek, this controller gender never has a distinctive morphology of its own. Rather, the agreement patterns make use of morphology from the other two genders. Such nouns are sometimes referred to as ambigeneric or ambigene, which is preferable to neuter, insofar as it refers to a phenomenon different from the neuter of Balkan Slavic and Greek.
Albanian, however, also has an additional ambigene pattern which, in traditional grammars, is labeled neuter. These are nouns that vary as far as their declension is concerned, declining like a masculine in the singular but with a plural definite article and declining like feminine plurals in the plural. Nouns in this class include substantivized participles (being replaced by feminine substantivized deverbal adjectives), a few nouns of adjectival origin, some mass nouns (pl in -ra, which requires feminine agreement, except fshat/fshatra ‘village/s’ which is ambigene in the plural) and krye ‘head’ (formerly also ballë ‘forehead’) in certain usages. This type of controller gender is becoming obsolete as items are made to conform to other patterns (feminine or ambigeneric) or are replaced by formations that so conform.Footnote 161
The origin of ambigenericity is to be sought in sound changes that led earlier neuters, from PIE in the case of Albanian and from Latin in the case of Romanian, to look like masculines in the singular and like feminines in the plural. While pressure to develop nouns in a particular way to give ambigenericity in Romanian and Albanian could date to a period of early contact between the languages, as suggested above for the unmarked status of feminine gender, the sound changes in question are language-specific. Moreover, Romanian ambigenericity is probably tied to a similar phenomenon seen to some extent in Italian (Rohlfs 1968: §370), where there are nouns like labbro ‘lip’ which is masculine in the singular, il labbro, but takes the feminine article in the plural, le labbra. Such facts would suggest that ambigenericity is an Eastern Romance development, and thus not specifically Balkan.
In most of BCMS, which, unlike Balkan Slavic, preserves gender marking in the plural, masculine animate nouns in /-a/, whose morphology follows that of a class of otherwise feminine nouns, are treated as feminine in the plural. In this respect, the Torlak dialects are transitional. There is competition between distinct and neutralized feminine plurals (-e and -i, respectively), but in either case, gender is consistent between the two numbers. Here the archaism of BCMS is striking, while Balkan Slavic can be compared to Russian in having leveled out plural gender agreement. It could be argued however, that both Russian and Balkan Slavic were subject to pressures toward simplification from other contact languages.
For the most part, then, gender distinction issues are language-internal. Although Latin had heterogeneous nouns with differing genders according to number, the particular pattern of masculine/singular + feminine/plural was lacking. Thus, the parallel in the specific categorial mergers in the realization of Albanian and Balkan Romance ambigeneric nouns might be part of the common heritage of Albanian and the language that shifted to become Romanian, even against the backdrop of possible broader Eastern Romance developments.
A gender-related phenomenon involving gender distinctions where contact is a likely cause concerns the Balkan Slavic third-person masculine-neuter singular dative weak object pronoun mu. In southwestern Macedonian, particularly the regions around Bitola, Lerin (Grk Flórina), Kostur (Grk Kastoria), Voden (Grk Édessa), and part of Prilep, this pronoun has expanded to replace the corresponding feminine form. According to Koneski 1981: 133, this replacement took place under the influence of Aromanian and Albanian (especially Aromanian in the case of Bitola), in which there is such syncretism in the dative singular clitic (cf. Sandfeld 1930: 120–121). Today the phenomenon is characteristic of the Macedonian spoken by Albanians in North Macedonia. Steinke & Ylli 2007: 318 report a similar phenomenon in Boboshtica, as well as the Prespa villages and Vrbnik (Alb Vërnik) (Steinke & Ylli 2007: 92, 266), and Golobrdo (Alb Golloborda) (Steinke & Ylli 2008: 82). All the Balkan languages have gender syncretism in the dative plural weak object pronouns.Footnote 162
6.1.3.2 Borrowing, Calquing, and Other Contact Effects Involving Gender
There are other scattered phenomena involving gender where the involvement of contact is not in question. Such is the case, for instance, with the treatment of borrowed adjectives in Romani. In its core vocabulary, Romani has two types of adjectives: inflecting (m.nom.sg -o, f.nom.sg -i, all other -e) and noninflecting (m.nom.sg in a consonant or borrowed adjectives in -o). In Arli varieties of North Macedonia, Macedonian adjectives are usually borrowed as invariants using their unmarked (neuter) form in -o, which corresponds to the Romani masculine, e.g., socijalno buti ‘social work,’ socijalno arka ‘social support’ (Romano Sumnal, Vol. 1, No 3, 94.04.01). Both buti and arka are feminine (the former native, the latter borrowed), and so should take agreeing adjectives in -i or -e depending on syntactic position. The following examples, however, show the effects of interference from speakers’ native Romani: buti normalni ‘a normal job’ (radio talk show, July 1994), kvalitetna evidentija ‘qualified documentation,’ privatikani karane ‘private reasons’ (instructions to census takers, June 1994). The first example shows a Romani feminine singular ending, the second a Macedonian feminine ending (the phrase is feminine in Macedonian), while in the third example the ending looks on the surface as if it is a Romani feminine singular, but in fact it is a Macedonian plural, influenced by the ending of the language from which the document was translated.
Also worth mentioning here are the effects of contact on gender in the creation of new verbal paradigms. Thus, for example, a key element of Gołąb’s argument (Gołąb 1984a: 7, 135; see §6.2.1) for the Aromanian origin of the ima ‘have’ perfect in Macedonian is the fact that the invariant neuter gender of the verbal adjective in Macedonian corresponds to the same invariant unmarked gender in Aromanian (the feminine); thus the Aromanian invariance is calqued with Macedonian’s invariant type. Similarly, Friedman 1994b and Markovikj 2007: 123–124 have argued that the shape of the base in the Frasheriote Aromanian admirative of Bela di Suprã (usually m.pl.impf) is influenced by the shape of the participle in the local Albanian dialects that served as the model (see §6.2.3).
Occasional gender shifts within a language can be attributed to influence from another language. A couple of examples, with words of importance to this study, are jazik ‘language’ and dijalekt ‘dialect,’ which appear as feminine in the Macedonian of Greece, as opposed to masculine in the Macedonian of North Macedonia, as a result of the influence of Greek; both γλώσσα ‘language’ and διάλεκτος ‘dialect’ are feminine nouns in Greek. The Macedonian speakers of Ajvato (Grk Liti) near Thessaloniki refer to their language as Našta ‘our.f.def.’ In general, pre-modern speakers of Macedonian would call their language našinski, naški, or some other masculine adjectival derivation reflecting, mutatis mutandis, naš jazik ‘our language.’ Leaving aside the Greek policies that have prevented these speakers from using the name used by other speakers of these same Lower Vardar Macedonian dialects (makedonski), the feminine form of the definite pronoun here is an excellent example of contact-induced gender shift.
Here we can also mention the fact that Eastern Diaspora Albanian dialects in Greece and Bulgaria and also Arvanitika have simplified gender agreement in the particle of concord. Thus, for example, the markedly feminine particle e has been replaced by të (or the functional equivalent thereof), which is used for various numbers, genders, and cases. This generalization has been arguably connected to the influence of the various contact languages (Bulgarian, Greek, Turkish) none of which have such gender/case/definiteness markers (Hamp 1965; Sokolova 1983: 60–61, 68; Sasse 1991: 301–303; Liosis 2021).
Finally, we can also note here the fact that the generalization of nonagreeing neuter adjectives is considered typical of nonnative (usually Turkish or Albanian) speakers of Balkan Slavic with the generalization of the least marked, i.e., neuter gender. Thus they produce BSl edno čaša ‘a cup’ (‘one.n cup.f’) instead of the expected edna čaša with feminine edna, or (Rudin 2019: Ex.(1)) Blg minala-to ljato i tozi ljato (‘last.f-the.n summer.n and this.m summer.n’) ‘last summer and this summer.’ This phenomenon is due to transfer from the speakers’ native Turkish, since Turkish has no gender at all, or, in the case of Albanian, the fact that Albanian has only two morphological genders. Such gender confusion also occurs when some Romani speakers use Macedonian, since their native language has only two genders. Koneski 1981: 132–133 notes a similar generalization of neuter attributives in some Lower Vardar dialects, e.g., mătno voda ‘troubled [muddy] water,’ kakvo čoek ‘what kind of person’ (where the adjective is n and the noun is m or f). These dialects would have been in intense contact with Turkish prior to the 1923 exchange of populations (see the relevant statistics in Kănčov 1900). Similarly, there are Macedonian nouns that have a gender that is unexpected compared to the usual shape for certain genders that pose problems for speakers of other languages; for instance, a small number of feminine gender nouns end in -C, the more typical ending for masculines, e.g., radost ‘joy.f’ (an old i-stem) and while some have shifted to masculine even for native speakers, e.g., život ‘life’ (only recently masculinized), those that remain feminine are a common source of mistakes for nonnative speakers.
6.1.4 Number
The category of number in the Balkan languages, like gender, is relatively restricted in terms of showing any significant contact phenomena. Borrowed plurals do figure in a matter of morphophonology, though that is discussed in §5.6. And, numerals show the effects of language contact, but they are treated as a lexical matter and discussed in §4.3.2. Accordingly, we focus here on a few developments in which number figures, the rather extensive borrowing and integration of the Turkish plural marker -lar/-ler into various Balkan languages, the adoption of plural markers in Romani, and the use of number in politeness.
6.1.4.1 Spread of Turkish Plurals
One key contact-related fact about the morphosyntax of number in the Balkans is that the Turkish plural marker -lar/-ler is found occurring on nouns in all of the languages, mostly, but not exclusively in words of Turkish origin. This fact was noted by Miklosich 1890: 9, who cited Serbian and Bulgarian agalari ‘agas; Turkish noblemen,’ adding Albanian agallarë ‘idem’ without comment. Sandfeld 1930: 161 uses essentially the same formulation but specifies Bulgarian, Serbian, and Albanian, giving at – atlar (StAlb atllarë) ‘horse – horses’ for exemplification.
The borrowing of inflectional affixes is interesting in its own right, as such affixes are generally held to be hard to borrow (Thomason & Kaufman 1988: 5–6, 14–16; and Matras 2009: 158, 212ff., who, however, suggests plurals as a recurring cross-linguistic exception); however, it is likely that Turkish -lar/-ler entered other languages attached to loanwords (see §6.1.4.1.5 below and cf. Gardani 2008; Seifart 2015). Moreover, the situation regarding this affix in the Balkans is considerably more complex than indicated by the brief treatments in Miklosich or Sandfeld, in terms of both the range of languages involved and the status of -lar/-ler in the various languages. For instance, in all of the languages, native plural marking can be or must be added. There is thus more to be said, and in the sections that follow, the facts from the individual languages are presented and evaluated.
6.1.4.1.1 South Slavic
South Slavic offers a particularly rich set of details concerning the use of the Turkish plural marker, as its occurrence in literary texts allows for some insight into the marker’s functions and connotations. Moreover, unlike many other discussions of Balkan features in this work, data from a broader range of BCMS prove relevant here.
6.1.4.1.1.1 Macedonian
Macedonian shows Turkish-based plurals to a considerable extent, in some instances with the Turkish suffix just on its own, e.g., kardašlar ‘brothers’ and in others with the Turkish suffix co-occurring with the productive native Macedonian plural ending, e.g., begleri(te) ‘(the) beys.’Footnote 163 Interestingly, too, in some cases as well the form in -lar/-ler is treated as a singular base for derivatives, e.g., Turče agalarče ‘O you young Turkish aga’ (Dimitrovski 1983: 8), with a diminutive suffix, or begler ‘bey,’ beglerka ‘bey’s wife,’ with a feminizing suffix. This is an important fact that suggests that -lar/-ler was not properly a plural marker in all instances in its occurrences in Macedonian. However, Grannes 1977: 90, citing Kowalski 1936: 13, argues that since plurals can be used with singular meaning in Turkish, especially as a plural of majesty, the reinterpretation of -lar/-ler as part of the stem came from a Turkish model rather than simple misunderstanding or reanalysis on the part of Macedonian speakers.Footnote 164
In terms of function, Jašar-Nasteva 2001: 215 notes that Turkish plurals occur almost exclusively as vocatives. She gives kardašlar ‘brothers’ (as above) in a context of direct address, and adds joldašlar ‘companions,’ efendiler ‘gentlemen,’ iplikler ‘clothes’ (lit., ‘threads’) without contexts.Footnote 165 This last example offers an interesting contrast with Albanian, which, as noted below (§6.1.4.1.2), reserves Turkish plurals exclusively for prestigious or important animate beings (almost all male human). A further function is that pluralized Turkish plurals are common in Macedonian toponyms, e.g., Jurumleri (older Urumleri), a village near Skopje (Jašar-Nasteva 2001: 182; Gołąb 1960). Finally, in parts of western North Macedonia with a large Muslim population, the Turkish plural suffix was also sometimes used to denote a group of family members using the first name of the relevant male, i.e., from the name Blaže the Macedonian plural would be Blaževci but in the aforementioned regions the form Blaželer also occurs (Grannes 1977: 84, citing Skok 1972: 578, cf. also §5.6).Footnote 166
6.1.4.1.1.2 Bulgarian
Turkish plurals in Bulgarian are particularly well documented. Grannes 1977 provides a wealth of examples, with extensive discussion, and includes twenty-four examples from literature (pp. 85–87). As in Macedonian, Bulgarian can treat such nouns as singular or plural and add the pluralizer -i to the form if its meaning is plural. Moreover, Bulgarian can also add the Slavic singulative suffix -in, e.g., aga – agalar ‘agas’ > agalari (pl) > agalarin (sg) ‘aga.’ All of Grannes’s examples refer to human males except at, which is usually translated ‘horse’ but also means ‘stallion,’ which is generally the meaning in the borrowing languages.
Some details of Grannes’s examples are quite revealing. The males are not always prestigious, as in his twelfth example on p. 86: De e čorbadžijata, bre, kjopolar! ‘Hey, where’s the boss, you sons of bitches!’Footnote 167 The speaker is a Turk and the addressees work in the shop of a non-Turk.Footnote 168 Of five abusive examples (kjopo[o]lar ‘son of bitches’ [twice], hainlar ‘traitors,’ kjaratalar (StTrk keratalar) ‘scoundrels,’ komitalar ‘rebel’ [abusive in context]), four are used by Turks, all to non-Turks, and one is used by Bai Ganyo (Konstantinov 1895) in an exclamatory comment on third parties. In fact, ten of the speakers in the twenty-one direct discourse examples are Turks.Footnote 169 An additional five are Turkish addressees for a total of fifteen out of twenty-one examples involving at least one Turkish interlocutor.Footnote 170 In only three examples is the Turkish plural used by a character (non-Turkish) in a narrative rather than a vocative context. The remaining eighteen are all vocatives.Footnote 171 The narrative uses all have the Bulgarian pluralizer -i added on, but no vocative use does.Footnote 172
Grannes, moreover, raises quite fairly the problem of distinguishing a codeswitch from a borrowing in this context (cf. §3.2.1.6).Footnote 173 Consider his first example (p. 85): Daj părvo na agite. Bujrjunus, agalar!, literally ‘give first to lord.pl.DEF. Please, lords,’ thus ‘Let the agas go first. Help yourselves, gentlemen.’ The speaker is Lazar (a Slavic speaker), and the first sentence is spoken to Vardarski (also a Slavic speaker), and the second to three Albanians. The first sentence is entirely in Bulgarian, including a native plural on the singular aga (agi). The second sentence is arguably a codeswitch into accented or dialectal Turkish (compare Bujrjunus with standard buyurunuz). Seven of the eighteen vocatives are of this type, i.e., occurring with a Turkish verb or an exclamation that could be Turkish (e.g., hajde ‘c’mon’). Of a similar type is the vocative çocuklarăm ‘my boys’ (example 5, p. 86), since the speaker and addressees are Turks. Thus, a total of eight of the seventeen vocatives are potential codeswitches.Footnote 174
Despite the richness of Grannes’s documentation of these Turkish plurals in Bulgarian, it must be admitted that very few if any of these remain in common usage in present-day Bulgarian, in part due to puristic pressures reshaping the standard language in the twentieth century. Still, their occurrence in nineteenth-century sources and in literature is Balkanologically relevant. See also Graham 2020: 99 for the eighteenth-century Bulgarian Paulician literature.
6.1.4.1.1.3 BCMS (“Serbo-Croatian”)
The occurrence of Turkish plurals in the Balkans is a topic where BCMS turns out to be of some, albeit limited, relevance. Popović 1960: 584 gives the examples hòđa-hođalári [sic] ‘Muslim cleric-s’ and hȁt-hatlári ‘stallion-s’ from Bosnia. However, as Grannes 1977: 84 observes, Serbo-Croatian belletristic literature, unlike that of Bulgarian, does not seem to use this formation. It does not occur in Bosnian Franciscan literature, examined by Graham 2020: 99. In his discussion of Turkish influence, Škaljić 1966: 11–62 makes no mention of plurals used as singulars, but he does have agalari from a Bosnian Muslim folk song recorded prior to World War Two. Other common words however, e.g., paša, beg, hodža, etc., are not cited as having Turkish plurals. He does have Hadžilar bajram ‘Festival of hajjis,’ also from Bosnia, but the singular hadžija has a plural hadžije, while the uninflected hadži does not mark number, so this is arguably a fixed expression. Knežević 1962 has no Turkish plurals aside from lexicalized examples such as tibȅlerosun ‘I don’t believe it,’ cf. Trk tövbeler olsun ‘I’ll never do it,’ tövbe ‘(vows of) repentance,’ ‘Enough!’.
Serbo-Croatian sources thus offer only limited testimony to the occurrence of Turkish plurals but a reasonable inference to be drawn from the older attested forms and the Bosnian provenance is that they were more prevalent earlier, despite their general absence from present-day usage, and that they were more common in areas with greater percentage of Muslims in the population, where Turkish lexical influence is stronger.Footnote 175
6.1.4.1.2 Albanian
Albanian affords an important perspective on the use of Turkish-based plurals. As in Balkan Slavic, in Albanian the Turkish plural marker is generally accompanied by a native plural ending, in this case the masculine animate -ë, added onto the Turkish one. It is significantly different from Balkan Slavic, however, in its treatment of Turkish plurals in that a small group of them are recognized as part of the standard language. Turkish plurals thus have a different sociolinguistic status in present-day Albanian from that of the corresponding forms in Slavic. Albanian also offers a glimpse into the chronology of Turkish loans since it is the case that none occur in Buzuku, the first long prose text in Albanian, published in 1555. This fact is perhaps also an indication of the sociolinguistic status of these loans in the early period, since Buzuku was a Catholic missal.
As for form, both -lar, spelled <-llar->, and -ler occur in these nouns, distributed according to the Turkish vowels in the source words.Footnote 176 Some of the Turkish-plural nouns in -a have alternative forms with retracted stressFootnote 177 in which the final -a is treated as a definite article, e.g., indefinite singular pashë ‘pasha,’ definite singular pásha vs. indefinite pashá, definite pashái, but both forms with the plural pashallarë. Fiedler 2007: 14–18 writes that a zero plural (pashë, agë) is also possible in the retracted forms, but this is not part of the norm. Moreover, -a can also occur as a plural marker in retracted forms, e.g., daí ‘maternal uncle’ (def daíu, pl dajllarë; Trk dayı but WRT dayi) or indefinite singular dajë – definite singular / indefinite plural daja. In Geg the indefinite plural of babë ‘father’ is baba. Although most authoritative sources claim that such plurals occur only with Turkish nouns – and that is indeed the rule in Standard Albanian – nonetheless a number of instances of -llar/-ler on non-Turkish nouns occur colloquially and dialectally in various parts of Albania and North Macedonia.
In the following two tables, a list is given, with bulleted notes, of nouns taking the Turkish plural in Albanian, first those in -llarë (Table 6.13) and then those in-lerë (Table 6.14). These forms are drawn from Kostallari 1976, the standard orthographic dictionary of the post-1972 unified standard, though they are all also in Cipo 1954, the dictionary that marked the solidification of a Tosk-based standard in Albania, as listed in Fiedler 2007: 314. Notes are added after each group on other nouns, taken from various sources, of which Fiedler 2007: 314–318 is the most important.
| -llarë |
| aga ‘noble’ |
| at ‘stallion’ |
| baba ‘father’ |
| pasha ‘pasha’ |
| hoxha ‘cleric’ (dialectal hoxhenj in Sulova, Haxhihasani 1955: 61) |
| subash ‘bailiff’ |
| usta ‘master craftsman’ |
| xhaxha ‘paternal uncle’ |
| dai (pl dajllarë) ‘maternal uncle’ |
| qehaja (pl in Cipo 1954 qehajallarë, in Kostallari 1976 qehallarë) ‘headman, overseer’ |
• Kaleshi 1971 cites shah ‘shah,’ çaush ‘sergeant,’ and hanëm ‘Turkish lady’ from North Macedonia and Kosovo, this last example being unusual in denoting a female, but a prestigious one; the prestige factor overrides gender, just as it does for species in at ‘stallion,’ the only nonhuman noun.
• Çeliku 1963: 207 also has komshi – komshillar ‘neighbor’ from north central Tosk; the singular here represents the WRT form, while the plural, although being added after final u > i, nonetheless harmonizes with the older vowel.
• Boretzky 1975: 237 also cites dost – dostllar ‘friend,’ mixhë – mixhallarë = axha – axhallarë = xhaxha – xhaxhallarë ‘uncle,’ baxhanak ‘wife’s sister’s husband’ with a plural baxhallarë from the hypocoristic stem baxhë; he also gives xhan – xhanllar ‘soul, beloved’ but with a definite xhanllari ‘beloved person’ (a singular, thus perhaps reflecting a misinterpretation).Footnote 178
| -lerë |
| dervish ‘dervish’ |
| efendi ‘gentleman’ |
| haxhi ‘hadji’ |
| kadi ‘judge’ (Fiedler 2007: 318 reports having heard the form kadillār in Zalladardhë near Peshkopi (where he heard kadilēr); he assumed the influence of StTrk kadılar) |
| sheh ‘sheikh’ |
| bejFootnote 179 ‘bey’ |
• Cipo 1954 gives only myftilerë as the plural of myfti ‘mufti,’ whereas Kostallari 1976 gives only myftinj. Both sources, however, have only kryemuftinj ‘head muftis’ as the plural of kryemyfti.
• Kaleshi 1971 also gives zabit ‘officer’ from Kosovo (zabitlerë vs. Kostallari 1976 zabitë; the word is absent from Cipo 1954).
• Kaleshi 1971 adds kollaxhilerë ‘monks’ (cf. StAlb singular kallogjer, from Greek καλόγερος).
• Boretzky 1975: 237: jediler ‘the seven,’ kërkler ‘the 40’ (only in fairy tales, where the number refers to a band of humans; note the lack of vowel harmony in ‘40’ [Trk kırklar]).
As noted above, Albanian extends the Turkish-based plurals to words of non-Turkish origin. Fiedler 2007: 317, for instance, cites the words mbret ‘king’ (from Latin) and prift ‘priest’ (also from Latin, though ultimately from Greek) as occurring dialectally with Turkish plurals mbretler, priftler, as opposed to StAlb mbretër, priftërinj. The former occurs in both East Central Geg and Central Tosk, while the latter is recorded for East Central Geg. Fiedler also cites the Greek-origin dhespot ‘bishop’ with a plural dhespotlerë, versus StAlb dhespotë, from Tahir Dizdari of Tirana, and trimllarë ‘heroes’ (singular trim, which is native Albanian) from an 1898 almanac. Çeliku 1963: 207 also recorded gjysh ‘grandfather,’ of native Albanian origin (cf. Orel 1998: s.v., < *sū-sa), with a plural gjyshllar in north-central Tosk territory. These forms show that -llarë/-lerë has taken on a life of its own, distinct from its Turkish roots. Nonetheless, all of these extended uses are in keeping with the prestige value associated with Turkish-origin nouns in Albanian with the Turkish plural affixes, inasmuch as ‘king,’ ‘priest,’ ‘bishop,’ etc. designate prestigious males.
In addition to all these nouns, it should be noted too that it is common colloquial practice in various parts of Albania (both north and south), as well as of North Macedonia and of Kosovo, for the Turkish-based -llarë/-lerë to be used for plural family or clan names, e.g., Brajkallarë, Cërnallarë, Durmishallarë, Bakijallarët, Haskallarë, Tafallarët, etc. (Nesimi 1987; Sobolev 2005c: 64; Fielder 2007: 317). Surnames such as Kostallari are of this same type.
Albanian thus shows the most extensive instantiation of the Turkish plurals of any of the contemporary Balkan languages, though there is evidence that the Turkish suffix has been reinterpreted in various instances.
6.1.4.1.3 Balkan Romance and Judezmo
For Aromanian, an early attestation of a Turkish-based plural is the form păshălarl’i ‘the pashas,’ cited in Leake 1814: 391. Polenakovikj 2007: s.vv. also gives agắ ‘lord,’ pl agalari (as well as less common agadzǐ), bimbaš ‘[army] major’– bimbašlar and beg – beglerańǐ, in addition to păšắ – pašalari. Although the Turkish plural is followed by a native plural marker for agắ, beg, and păšắ, bimbaš is cited with only the Turkish plural marker. Meglenoromanian has the form păšălărés as an attributive adjective from pášă ‘pasha,’ which seems to be formed from the Turkish plural (Capidan 1935: s.v.). For Judezmo, Varol Bornes 2008: 211 recorded the form askerleres ‘soldiers’ (asker + Turkish plural + Judezmo plural) .
6.1.4.1.4 Greek
The absence of Turkish plurals in Greek is stated explicitly by Kyranoudis 2007: 147–148. Even the most heavily Turkish-affected dialects in Asia Minor (Cappadocian and Pontic) seem not to have adopted any Turkish plurals. However, Joseph 2016a reports on the occurrence in the Greek of southern Albania (specifically Saranda (Grk Ágioi Saránta)) of the form xodzaláres ‘Muslim clerics/teachers’ as the plural of the loanword xodzas (Turkish hoca). This form shows the Turkish -lar plural plus the productive Greek plural ending -ες. The form was produced by a Greek-dominant Greek-Albanian bilingual man, and while it does indeed have the Turkish -lar- in it, it most likely does not represent a direct borrowing of the Turkish plural but rather is directly built on the Albanian hoxhallarë (cited above in Table 6.13). Similarly, a further instance, αγαλάρδες ‘agas’ has been reported by Aristotle Spirou (p.c., June 2016), from the same general area (Himara, c. 50 km north of Saranda). It too shows the Greek -ες, though added to an extended stem in -δ-, an element added to many loanword plurals, as in standard Greek χοτζάδες, the plural of χότζας, and even some to native Greek words, e.g., πατέρα- ‘father,’ with a familiar plural πατεράδες ‘dads.’ Limited though they are, such forms show how such foreign plural material can become embedded in a language.
6.1.4.1.5 Romani
Although Romani has borrowed plural markers from some of its contact languages, e.g., Greek, Romanian, to some extent Slavic, Turkish plurals are uncommon. Nonetheless such forms do occur on occasion, e.g., beglerja ‘beys,’ pašalarja ‘pashas,’ where -ja is a Romani plural marker (Cech et al. 2009: 220, 230). In some dialects of Romani under heavy Turkish influence, the 3pl pronoun on becomes onlar (sometimes onnar). As it so happens, the plural of the Turkish neutral 3rd person pronoun o ‘he, she, it, this, that’ is onlar, and this surely is part of the borrowing process. We can also mention here the use of -lar ~ ler as a 3rd plural marker on Turkish verbs in those Romani dialects with Turkish conjugation (see §6.2.1.1.1 below).
6.1.4.1.6 -lar/-ler Summation
The appearance of the Turkish plural ending in all the Balkan languages, albeit to varying degrees, is striking and seemingly of particular morphosyntactic interest for Balkan linguistics. Moreover, it attests to the importance of Turkish in the Balkans during the Ottoman period. Still, in a certain sense, it is not all that remarkable a development, and while certainly one that is worthy of mention, it perhaps does not occupy an unusual place in Balkanology. The fact that it occurs mostly on Turkish loans makes it clear that it entered the languages attached to these loans and spread, if at all, from there to non-Turkish words. In that way, it simply reflects lexical borrowing, though admittedly of the somewhat rare type in which a fully inflected form is borrowed and the inflection, sometimes, moves on. When reinterpretation and spread occur within the borrowing language, it is no more special than the spread of Latin plural markers in colloquial English, e.g., octopus/octopi. A difference between the Balkan phenomenon and English is that while, e.g., datum/data is a learnedism (some might even say an affectation), the Turkish plurals in the Balkan languages are a part of folk speech. However, it is entirely possible that Turkish plurals entered the Balkan languages via speakers who were showing off their knowledge of the language of power and prestige.
6.1.4.2 Other Plural Inflections
In a few of the languages, and thus on a far less extensive basis than what is seen with the Turkish plural marker (§6.1.4.1), plural markers occur that are borrowed from other Balkan languages.Footnote 180 Several instances, involving different source languages and different plural markers, can be cited from Romani, though that is not the only language affected in this way.
In particular, Vlax dialects of Romani show a plural marker -urja (North) and -ura or -ujra (South), as in, e.g., Vlah-urja ‘Vlachs’ (Boretzky & Igla 1994: 398), which has entered Romani from Romanian, specifically drawing on the Romanian plural suffix -uri, as in tîmp-uri ‘times.’ Further, Sechidou 2011: 30, writing about the Romani of Ágios Athanásios in Greece, cites instances of o-stem nouns (nominative singular in -os) with a plural ending -mata, e.g., kumbaros ‘best man’ / kumbaromata ‘pl’, izos ‘thing’ / izomata ‘pl’, zjaros ‘burning coil’ / zjaromata ‘pl.’ This marker comes from one class of Greek neuter plurals, specifically from nouns with a singular in -μα that have a nominative/accusative plural in -ματα.Footnote 181 Some Romani -ma nouns have been borrowed as such, e.g., namborema ‘illness,’ from Greek ανημπόρεμα, but the -mata ending also occurs in Romani on nouns that do not have this plural in Greek – although kumbaros is a borrowing from Greek, in Greek it is masculine and has a nominative plural in -οι (κουμπάροι) and accusative plural in -ους (κουμπάρους) – and on nouns borrowed from other languages, e.g., zjar-, a likely borrowing from Albanan zjarr ‘fire’ (whence it has entered some northern Greek dialects, Meyer 1891: s.v.; cf. also BSl žar ‘grill, požar ‘[wild] fire’).Footnote 182 This Greek suffix also occurs in the dialect of Parakálamos (Matras 2004: 74). Dialects of Romani in contact with Slavic show plurals in -i and -ovi, especially in North Macedonia (Boretzky & Igla 1991: 32; Cech et al. 2009: 214–231).
Finally, the Greek suffix -δες, consisting of the stem-extension -δ- that occurs with some plurals (so-called “imparisyllabics,” cf. Holton et al. 1997: §2.1.5 and §6.1.4.1.4) and the productive ending -ες,Footnote 183 occurs in various Balkan languages. Romani dialects of the Balkan II type (Boretzky & Igla 1994: 379) show it, some in words borrowed from Greek, e.g., kafés ‘(cup of) coffee’ / kafedes ‘pl,’ but also in other words as well, e.g., Bugurdžis ‘Bugurdži’ / Bugurdži-des ‘pl,’ paša ‘pasha’ / pašades ‘pl.’ A variant ending -da occurs, an apparent contamination of -des “mit der indigenen Pluralending -a” (‘with the indigenous plural ending -a’) (ibid.). Moreover, Sechidou 2011: 31–32 observes for Romani of Ágios Athanásios that plurals in -as/-es/-is occur, from Greek -δες with different stem-thematic vowels, as suggested also by the variants -aes/-ees/-ies. Examples include dajf-as ‘relative’ / dajf-aes ‘pl,’ komš-is ‘neighbor’ / komš-ies ‘pl,’ where Turkish-origin nouns are involved. Interestingly, though it is not clear if this is a contact-induced effect or just an independent development in each language, both Ágios Athanásios Romani and Greek show “a tendency for the expansion of the plural suffix” with theme vowel -a- at the expense of the -i-based allomorph (Sechidou 2011: 32). Thus Romani has čandirdž-aes as the plural of čandirdž-is ‘Romani group designation’ (instead of expected čandirdž-ies), much as Greek has, as Sandfeld 1930: 104 noted, e.g., βουλευτ-άδες ‘deputies’ as the plural of βουλευτ-ής (for expected, even if no longer current, βουλευτ-ήδες).
And, this Greek suffix -δες is found in Balkan Romance as well. Sandfeld 1930: 104–105 observes that Aromanian has a plural ending -dz(i)/-ts(i), as in cãsãbã / cãsãbadz(i) ‘town/s’ or udã / udats(i) ‘room/s,’ kafe / kafedz(i), which is from Greek -δες, and Meglenoromanian shows the same Greek marker, in, for instance, kafegij ‘coffeehouse proprietor’ / kafegiaz ‘pl,’ where -z reflects -δες (cf. Capidan 1925a: 147).
6.1.4.3 Number and Politeness
The use of pronouns and verbs in the plural serving as a polite singular – the so-called T/V distinctionFootnote 184 – is a relatively recent phenomenon in all of the Balkan languages, as noted also in §6.2.4.3.4.2. Hints of such a distinction occur in Late Latin imperial usage and it spread across Western Europe court usage in the Middle Ages (Brown & Gilman 1960: 106–107). Yet, in the oldest Romanian document – the 1512 letter of Neacşu of Câmpulung – the T forms predominate overwhelmingly, despite the fact that the letter is addressed by a vassal to his lord (warning the latter of an impending attack). Thus the ‘you’ in ‘your highness’ is a T form (do[m]nïjata, do[m]nïetale) in nine occurrences plus one occurrence of la tine ‘to yourself’ and V in two (domniile vo[s]tre). The verbs are in the second person singular in six occurrences (all štïi ‘know’) and in the second person plural in one (ešti ‘are,’ used with do(m)nïjata), and the closing, which is in Wallacho-Bulgarian (i bog te veselit ‘may God make you happy’), likewise uses a T form (te). On the evolution of the deferential forms of address in Romanian, see Uța Barbulescu 2019.
During this same period or earlier, such usage did not occur in Slavic, where, for instance, the T/V forms marked only singular/plural in Medieval Rus,’ nor in Albanian. Such usage seems thus to have spread from Western Europe into the various Balkan languages (and if found in Judezmo, it could be an inheritance from Late Latin). In Romani, it has only begun to occur in the post-1989 rise of Romani linguistic and political movements, and the distinction in Macedonian is basically post-World War II with speakers today showing considerable variability. In Greek, it is mainly a feature of urban usage, and in some rural areas, e.g., parts of southern Albania, singular forms are regular in address to foreigners, where urban usage would dictate a plural of politeness.Footnote 185
This phenomenon is thus a feature found in the various languages now but it has its origins in modernization and westernization in the Balkans. As such, as interesting as it is in general and as important as it is for the sociolinguistics of the present-day Balkans, it has no bearing on the Balkan sprachbund.Footnote 186
6.1.5 Adjectives – Gradation
Adjectives in a sense present the least complicated area for consideration in a presentation of Balkan morphosyntax.Footnote 187 While Balkan language contact has allowed some anomalies to enter the adjectival systems of some languages, such as uninflected adjectives like Macedonian taze ‘fresh,’ from Turkish, or Romani normalno ‘normal,’ from Macedonian, as discussed in greater detail in §6.1.3.2, such cases are more a matter of contact-induced lexical particularities for the recipient languages and significant widespread systemic morphosyntactic developments.Footnote 188 Accordingly, we focus here on the one area of adjectival morphosyntax where contact does seem to have played a role in the Balkans, namely gradation.Footnote 189 We see adjectival gradation as an inflectional issue and thus a matter of morphosyntax and not lexicon per se, even if in some instances, the lexical material involved in marking the categories is borrowed. Cross-linguistically there are several gradations for adjectives, but the main ones of concern here are the comparative and the superlative (but cf. also §4.2.2.3).
A key fact about developments with comparatives in the Balkans is that in all of the languages, they show movement from synthetic structures to analytic ones. Synthetic gradation survives at the peripheries, as they still constitute a viable formation in Greek, especially for comparatives, and in northern Torlak BCMS, where a small number of such forms can be found, and Romani dialects outside the Balkans are also more likely to preserve synthetic comparatives. We can also note here that in some southern Montenegrin dialects, where the analytic superlative marker naj- is normally added to the synthetic comparative, the comparative marker po- can also be added to the synthetic comparative to mean ‘much more than others’ (Stevanović 1935: 80–81). We see in this a sort of midway point between true analytic and synthetic comparatives. The superlatives present a somewhat more complicated picture. Still, adjectival gradation overall is consistent with other Balkan developments showing movement towards analytism.
Nevertheless, there are various considerations that need to be taken into account in assessing the Balkan situation with gradation. For instance, analytic structures are part of Slavic inheritance for superlatives, inasmuch as they are found throughout the Slavic family. Yet, there is a particular Balkan instantiation to the realization of superlatives in Balkan Slavic. Moreover, Romance and Germanic show similar changes with gradation in the direction of analytic structures (cf. Frn plus intelligent / Eng more intelligent), so there may be something natural or typological about this kind of development. (On the other hand, English analyticity tends to be more with Latinate words.)
The basic facts on analytic gradation are given below in Table 6.15, exemplified with ‘big’ in each language, thus ‘bigger’ and ‘biggest.’
This basic picture of the facts, as presented, calls for some elaboration and various clarifications.
Table 6.15 shows that in the superlative, Turkish and Balkan Slavic represent one end of a scale, with Greek and Albanian at the other end, and Balkan Romance and Romani occupy the middle ground. That is, both Turkish and Balkan Slavic have a purely analytic relative superlative and use native markers: Turkish en, BSl naj. In Greek, including Thracian Greek, and Albanian, on the other hand, the definite of the comparative is the basis for the relative superlative, using entirely native material (though see §6.1.5.3 for some discussion of Greek πιο and below on Turkish influence on κομ). Romanian, along with most of Meglenoromanian, shows a pattern like that of Albanian, utilizing native marking, whereas Aromanian and the Meglenoromanian of Tsărnarekă have borrowed Slavic naj. Romani and the Greek of Thrace (Kırklareli, Grk Saránta Ekklisies) and of Southern Albania (Palasë, Grk Palása) show analytic forms marked with material that is borrowed outright (e.g., Alb ma – Joseph et al. 2019: 238) or, in the case of κομ, calqued (κομ from ακόμη ‘still,’ calqued from Turkish daha ‘more’ but also ‘still’).Footnote 190
Several important additional sets of facts are not evident from Table 6.15. These are treated, language by language, in the sections that follow.
6.1.5.1 Balkan Slavic
There is great consistency across Balkan Slavic with adjectival gradation. Like Bulgarian and Macedonian, Torlak and Southern Montenegrin BCMS dialects show analytic comparatives with po and naj.Footnote 191 The superlative formation with naj plus an inflected comparative is found occasionally in OCS, though more usually the comparative form is simply used for the superlative. Analytic gradation is thus part of the inheritance from Common Slavic into South Slavic and ultimately Balkan Slavic. Still, noteworthy from the Slavic point of view is the fact that in the superlative, naj is added to a bare adjectival form (Blg naj-dobăr, Mac najdobar ‘best’), whereas for the most part, the rest of Slavic adds naj to the inflected comparative, e.g., Ukr najslavshij ‘weakest’ (cf. slavkij ‘weak,’ slavshij ‘weaker’).Footnote 192 The Bulgarian and Macedonian superlative formations suggest a Balkanization of adjectival degree inflection in Bulgarian and Macedonian, though the absence of an overt comparative suffix here could also be attributed to the general Balkan Slavic loss of grammatical inflections outside of the verb.
On a lexical level, but worth noting here as it incorporates a foreign inflectional marker, is the Turkism beter ‘worse, horrible’ from Persian badtar ‘worse,’ the comparative of bad ‘bad.’ The term is considered archaic for Bulgarian (Grannes et al. 2002: 30) and is old-fashioned colloquial in Albanian, but it is very much alive in Macedonian, where it is not only common in colloquial usage but is also deployed by speakers who know English for humorous purposes. We can also note here lexical survivals like Mac povekje, Blg poveče ‘more,’ which function as comparatives of mnogu/mnogo ‘much, many, etc.’ (see §6.1.5.5).
6.1.5.2 Balkan Romance
The starting point for Balkan Romance, Latin, had synthetic gradation as the norm, but analytic types were available, a comparative with magis ‘more’ and a superlative with maximē ‘most,’ both preceding the positive form. Some adjectives allowed both, e.g., ēlegāns ‘select’ formed both comparatives ēlegantior / magis ēlegāns and both superlatives ēlegantissimus / maximē ēlegāns. The analytic type no doubt prevailed for the comparative in Vulgar Latin to judge from the fact that this pattern is found across all of Romance.
Balkan Romance continues the analytic marking of the comparative, with mai, from magis, in Romanian and Meglenoromanian, and cama, composed of ca ‘as; that; how,’ from Latin quam ‘than,’ plus ma ‘more; but; again,’ from magis. The Aromanian of Metsovo, in Greece, uses both cama and mai (Beis 2000: 400). In the superlative, the pattern with definitivization of the comparative is found elsewhere in Romance (cf. French la plus jolie ‘the prettiest,’, lit., ‘the more pretty’), so even though both Greek and Albanian have it, it need not be a contact-induced feature in Balkan Romance. Gołąb 1984a: 85 cites nai from Slavic in the Aromanian of Kruševo, e.g., náibűnu ‘the best’ (positive bunu, comparative káma bunu), with the Slavic pattern of building the superlative on the positive form. Similarly, Markovikj 2007: 68 cites naj as the prefix for Frasheriote Ohrid Aromanian. The Slavic intrusion itself is of interest, but Gołąb observes further that nai “can be treated as a particle (prepositive) since it clearly bears a secondary stress, and besides that it can be also used with compound adverbs or adverbial expressions like e.g., nái-tu-sőne ‘at the very end, at last’” (with tu ‘in’ and sone ‘end’); see more discussion in §6.1.5.6. For emphatic purposes, Frasheriote Ohrid Aromanian can have forms of the type najcama bun (Markovikj 2007: 68), with the more usual Balkan Romance pattern of building the superlative on the comparative, but realized via the Slavic-origin marker; some other dialects make the comparative definite as in Albanian.
6.1.5.3 Greek
In addition to the analytic pattern seen in Table 6.15, Greek has synthetic forms in the comparative with the suffix -τερος that remain in competition with the innovative analytic patterns. Thus one finds both μικρότερος (cf. μικρός ‘small’) and πιο μικρός ‘smaller,’ ομορφότερος (cf. όμορφος ‘beautiful’) and πιο όμορφος ‘more beautiful,’ etc. Not all adjectives can form both (see Holton et al. 1997: 86–87 for discussion) but both patterns are quite viable and alive. Such a statement is less true for the synthetic superlative; there is a superlative in -τατος, with the Ancient Greek suffix, e.g., βαθύτατος ‘very deep indeed, profound,’ φυσικότατος ‘most natural; entirely natural,’ but it is used as an absolute superlative, emphasizing a particularly noteworthy quality, not as a relative superlative (involved in a direct comparison against a certain set of comparanda) in the way that the analytic type is used. Moreover, as Holton et al. 1997: 88 describe it, “the absolute superlative is a rather learned form: it is much less frequently used than the comparative and the relative superlative and cannot be produced for all adjectives.”
Still, the synthetic patterns reflect Ancient Greek means of marking adjectival gradation, and the modern suffixes are inherited. Interestingly, as with Latin (cf. §6.1.5.2), Ancient Greek had analytic patterns with μᾶλλον ‘more’ and μάλιστα ‘most’ preceding the positive form, so there was a precedent internal to Greek for the analytic type. Moreover, Ancient Greek πλέον ‘more’ (neuter singular of πλείων ‘more’) could also be used in the way that μᾶλλον was, e.g., by Thucydides (Smyth 1920: §1068). However, after a careful review of the data from the Hellenistic-Roman period, Markopoulos 2015: 212 concludes that this pattern, especially with πλέον, “seems to have been short-lived … [and] did not survive long” past Hellenistic-Roman times. Thus, the modern construction may well be an innovation. Markopoulos 2015: 226 points to Romance as the source of the analytic formation with πιο, saying it was “the result of language contact between Greek and Romance-speakers in the late Middle Ages (ca. twelfth to sixteenth centuries).” And, the analytic superlative could very well be a conscious importation from Romance (perhaps French) during the nineteenth century. The usual etymology (cf. Andriotis 1983: s.v.; Babiniotis 2010: s.v.) takes the marker πιο to be from earlier πλέον, and Thumb 1912: §119.1 and others do give variants with -λ- (πλεο ([pljo])),Footnote 193 but the similarity to Romance forms, especially Italian più, is suggestive of borrowing.Footnote 194
6.1.5.4 Turkish and Turkic
The expression of analytic adjectival gradation is old in Turkish, attested in the oldest monuments, those from the eighth century. And it is clear that Turkish was a key source of material pertaining to the expression of comparatives and superlatives in Romani and Thracian Greek. It is an open question, however, if the role of Turkish extended beyond the lexicon and if Turkish played any part in the structural changes giving analytic patterns; some have thought this possible (Georgiev 1952) while others consider it unlikely (Schaller 1975).
Although the evidence of Late Latin and of pre-Ottoman Greek and Balkan Slavic give indications of early analytic structures in those languages (but cf. §6.1.5.3 on the demise of a Hellenistic Greek analytic comparative), one might suggest that Turkish could have contributed in terms of both feature selection, adding to the pool of support for analytic formations, and expansion. The degree to which analytic structures replaced the synthetic ones increased significantly precisely during the period of intensive contact with Turkish. In the case of Slavic, only those dialects south of the northern boundary of Torlak BCMS show the complete elimination of the old synthetic comparative.
As a related Turkic point, we note that in Moldavian Gagauz, sam (< Russ samyj) is in competition with en as the superlative marker for the younger generation of speakers; the analytism remains but with a new element competing.
6.1.5.5 Romani
Romani presents a more complicated and richer picture than Table 6.15 indicates. Elšík & Matras 2006: 146–155 provide a detailed account of adjectival gradation in a large sample of Romani dialects as do Boretzky & Igla 2004: 1.184, 2.78n,78s. The inherited comparative in -eder (cf. Skt -tara-, PIE *-tero-) is preserved with great consistency in those dialects that left the Balkans early and became the Central and Northern groups (Boretzky & Igla 2004: Map 78n).Footnote 195 The dialects that became the Vlax group, so named due to time spent in contact with Romanian (see §1.2.3.5), have eliminated almost all vestiges of the inherited comparative marking, replacing it with maj, from Romanian mai. Maj is retained in these dialects despite later migrations either within the Balkans or beyond. One exception is the Romani of Agía Varvára, in Greece, in which Slavic-based pa and Turkish daha occur (Boretzky & Igla 2004: 1.184, 2. Map78s).Footnote 196 The Slavic markers po and naj and the Turkish daha and en (including WRT variants daa, em as well as taa, thaa, and xen – Elšík & Matras 2006: 151) occur in a wide variety of combinations throughout the Romani dialects of the Balkan group: po alone, po naj, po naj en, po daha, po en, po daha en, and daha en. At least one South Vlax dialect in north Bulgaria has maj, po, naj, and eni. Vlax maj, besides its resilience within Vlax, has spread to a few Balkan dialects, e.g., Spoitori, a South Balkan II dialect in Romania, where po competes with it, and a South Balkan I dialect in Crimea, where it has replaced po. Given that Romani entered the Balkans sometime between the tenth and thirteenth centuries, and given that during this same period Slavic preserved its inflectional system of adjectival gradation, it would appear that Balkan Slavic and Romani were undergoing this shift at about the same time, and those dialects that left the Balkans did so before its completion.
It is worth noting, however, that even in the Balkan dialects, which generally show strong convergences with their co-territorial languages in analytic gradation via borrowing and calquing, individual lexical items sometimes preserve the older synthetic comparative, e.g., but ‘many, very’ > buteder ‘more’ or pobuter. This occurs in North Macedonia and Bulgaria, where the respective nation-state languages have a suppletive comparative (inherited from an older synthetic comparative) in the same lexical item: mnogu/mnogo ‘many, very’ versus povekje/poveče ‘more.’
Also significant is the fact that some Romani dialects follow the pattern of the current co-territorial nation-state language pattern in using the definite comparative to form the superlative (as in Romanian, Albanian, and Greek), e.g., Agía Varvára near Athens still uses the Turkish or Slavic form of the comparative in forming the superlative, e.g., daha/po baro ‘bigger,’ but behaves like Greek in using the definite comparative to form the superlative o daha/po baro ‘the biggest,’ likewise Florina Arli (e.g., po-baro, o po-baro ‘idem’), and a similar pattern occurs in Prizren Arli based on Albanian (Elšík & Matras 2006: 153). An exception is the Arli of Kardítsa in northern Greece, where older Macedonian po is retained for the comparative but pio is borrowed from Grk πιο ‘more’ to mark the superlative (versus definite + πιο in Greek; Elšík & Matras 2006: 152–153).
6.1.5.6 Miscellaneous Differences
The individual languages differ to some extent in how the analytic structures are realized for adjectives. In all of the formations given in the sections above, one can see analytism because the markers had separate word status at some point or show characteristics associated with words. But in Slavic especially, the markers may be best thought of as bound words, ones that do not ordinarily occur independently, except for expressive purposes, usually reduplicated. Moreover, in some of the languages, in particular Bulgarian (though not Macedonian) and at least for naj, they have their own accent, resulting in formations that might be best characterized as compounds rather than phrases, e.g., náj-dóbăr ‘better’ (moreover, Bulgarians need to be taught to write the dash, as they perceive naj as a separate word); also, note what Gołąb (using somewhat different terminology) says about Aromanian náibűnu, above in §6.1.5.2. And Standard Albanian më (Geg mâ), even though it is bound to its adjective, does require the particle of concord (if there is one) to intervene between it and the adjective, and moreover, does not show any vowel contraction with the particle. For instance më e bukur ‘more beautiful.f’ does not contract to *ma bukur, unlike the combination of homophonous weak object pronouns (1sg.dat më + 3sg.acc e) which do contract to ma, as in ai ma mori ‘he got it for me’ (‘he for.me.it got’), and unlike the behavior of the subjunctive marker (dms) të, as in do ta marr ‘I will get it’ (fut dms.it get.1sg). These facts suggest a looser connection between comparative më and what follows it than with other seemingly similar combinations.
Another interesting feature of these markers is that in some of the languages, on a limited basis, they can attach to some words that are not adjectives. In Greek, Balkan Slavic, and Romani, for instance, the comparative markers can attach to and modify nouns, e.g., Slavic po-junak ‘more a hero,’ Grk πιο άνθρωπος ‘more a man,’ Arli Rmi po-budala ‘more of a fool.’ Bulgarian, Macedonian, Torlak BCMS (A. Belić 1905: 438–440), and some Romani (Boretzky & Igla 1994: 383) go even further in this regard and allow po and naj to modify verbs, e.g., Blg po-xaresvam ‘I like (it) better,’ Mac naj ne sakam ‘I dislike most of all,’ Bugurdži Rmi po-džanla ‘s/he knows more/better.’ This last type of usage is reminiscent of, but not related to, Classical Sanskrit pacatitaram ‘cooks very well, cooks better than someone else’ (cf. pacati ‘cooks’, -tara- ‘-er’); such a formation is impossible in Greek (*πιο αγαπάω ‘I more love (it) than (feel) anyway else’).
Differences of these sorts are to be expected for seemingly similar phenomena even within zones of convergence; similarity in structure does not mean identity for all elements in all respects. For instance, the accentual adjustment required by extended prosodic domains in light of the three-syllable-window for stress placement differs in realization between Greek and Macedonian, so that the phenomenon of “double accent” is different across the two languages (see §5.5.2); the indefinite article derived from ‘one’ in all the Balkan languages has different properties in each language (see §6.1.2.4.1); and multiple-WH-fronting structures in Balkan and non-Balkan Slavic differ in configurational detail even though they are linearly the same (see §7.8.3.2).
6.1.5.7 Summation on Gradation
The chronology of beginnings of analytism and widespread nature of analytic structures outside the Balkans make it seem like not the most likely candidate for a contact-related account, though the sort of scenario sketched, for example, for the rise of prepositional (i.e., analytic) marking of indirect objects in §6.1.1.1.3.2, or for the infinitival developments (also analytic) in §7.7.2.1.4, would seem to be applicable here. That is, there seems to be a suitable mix of pre-existing tendencies in the languages, as well as no obvious evidence pointing to one language as the instigator. Thus, mutually reinforcing multilateral contact allowing for selection of similar variants from a pool of possibilities offers a reasonable account of what may have happened. Additionally, the very real lexical effects seen in various of the languages show that this was a domain of grammar where contact played a role. Moreover, to the extent that these markers for gradation in Turkish are grammatical and not just lexical, i.e., inflection that is realized periphrastically, the borrowings show, once again, that grammatical elements can pass from one language into another.
6.1.6 Nominal Analytism – The Matter of Causation
As with analytism specifically in Balkan case marking systems – see §6.1.1.6 – the question of causation for the rise of analytism in general in the nominal system must be addressed. While for case systems, the interplay of particular languages provided an insight into why the developments unfolded as they did, for analytism more generally, it would appear that the nature of language contact itself is a major contributory factor. It is well known that language contact favors analytic structures in that the communicative needs of interlocutors are better met with analytic structures that allow for easier parsing by listeners and thus better understanding. And in responding to questions or comments, or just keeping up their end of a conversation, a listener-turned-speaker would for the most part be able to choose among variant modes of structuring a reply, thus selecting among variants in such a way as to accommodate to the level of proficiency of the interlocutor in the language being used as the medium of communication. If analytic structures are among the variants available to both speakers – and it is important to note that for all the Balkan languages there were structurally synthetic and analytic counterparts that speakers could draw on – then the selection of analytic forms would have facilitated communication. Thus Balkan language contact would quite naturally have turned the languages in the direction of greater analytism in structure.
6.2 Verbal Morphology and Morphosyntax
In discussing Balkan contact phenomena in the verb, we can distinguish five general types of features: (1) borrowed or copied (Gołąb 1976: 304; Johanson 1992) morphemes replacing or added to older morphemes, (2) calqued constructions replacing older constructions, (3) new categories built on native material, (4) new categories constructed using borrowed/copied material, and (5) convergences using native material. Section 6.2.1 illustrates the first and fifth types of changes in tense-person markers, on the one hand, and absolute versus relative tenses and simplex versus analytic pasts on the other. Section 6.2.2 on aspect, including superordinate and subordinate aspect, aktionsart, and intersections of aspect and mood, shows all five types of changes, and various languages as the sources, though Slavic is most important for aktionsart. Section 6.2.3 on taxis and resultativity focuses mainly on the rise of new perfects and pluperfects, for which the second, third, and fifth types of change are most important. Here Latin or Balkan Romance plays a particularly important role. Section 6.2.4 is concerned with futures, conditionals, and volitionals such as optatives, imperatives, and prohibitives, and here the second, third, and fifth types of change are again most prominent; still, Greek models seem to have been particularly influential, although the narrative imperative (see §7.8.2.2.8) appears to be a convergence based on a Slavic model. Section 6.2.5, on evidentiality, involves the third and fourth types of change, but also the fifth. While Turkish is undoubtedly the initial impetus, subsequent influences from Albanian onto certain Aromanian and Macedonian dialects, Macedonian on Meglenoromanian, Balkan Slavic on Romani, and Romanian on Novo Selo Vidinsko Bulgarian are also all relevant. Section 6.2.6, on voice and valency, is mainly concerned with convergences connected with lability, although the influence of Macedonian on Romani and dialectal Greek impersonals is also relevant here.
6.2.1 Tense
Tense/aspect systems tend to be more resistant to contact-induced change than modal systems. Thus, for example, the basic tense/aspect system of Romani – whose adult speakers are all bi- or multilingual in a broad range of languages – is extremely stable pan-dialectally, whereas the Romani modal system is always calqued or borrowed. Based on a broadly cross-linguistic study of contact-induced change, Matras 2007: 44 writes: “Little attention has been granted in the literature to the borrowing of features belonging to the domain of verbs … ; reports on the borrowing of TAM markers are quite rare.” He posits the hierarchy of borrowability given in (6.70):
In the Balkans, if one treats the replacement of infinitives with analytic subjunctives as a contact-induced modality phenomenon, one nevertheless cannot identify an equivalent aktionsart/aspect calquing of similar cross-linguistic and geographic extent, although localized phenomena do occur. The shared creation of a future based on a particle descended from a verb meaning ‘want’ in various Balkan languages, however, has basically the same distribution as that of infinitive replacement by an analytic subjunctive, thus contributing to the argument for treating futurity as a mood rather than a tense category (as was argued by Kuryłowicz, 1956 and Gołąb 1964b). We return to this problem in §6.2.4.1 below (see also Joseph 1983a). Matras 2007: 45–46 argues that the less stable, secure, or intimate the event is from the speaker’s perspective, the more likely the verbal category is to be borrowed. He treats mood as denoting speaker control, aspect as denoting the internal structure of the event, and tense as denoting the most intimate relationship of the event to the speaker’s perspective. We can note here that futurity can also be treated as modal in this sense.
If we compare these definitions to those given in Jakobson’s 1957 classic article on shifters and verbal categories, namely mood as the relation PnEn/Ps, aspect as the quantification of En, and tense as the relation EnEs (where P = participant, E = event, s = speech, and n = narrated), we see that Matras’s cline of likelihood of copying (to use the terminology of Gołąb 1976 and Johanson 1992), makes the category that includes the speech event (i.e., the shifter) more likely to be borrowed, but the nonshifting category (aspect) is between two shifters. If, however, we accept Aronson’s 1991 argument, based on Gołąb 1964b, that mood represents the ontological evaluation of the narrated event, i.e., the qualification of En in Jakobson’s terms, then the one of these categories that is a shifter, i.e., tense, is the one intimately connected to the speaker in Matras’s terms.Footnote 197 Our approach here is informalist, to use Binnick’s 1991 felicitous term, and for us the basic tense opposition is past/nonpast.
6.2.1.1 The Morphology of Tense–Person Markers
The correlation or combination of person marking with tense marking is characteristic of inflecting languages, but even Turkish, a classic agglutinating language, shows some conflation of tense and person, e.g., -k for1pl aorist/conditional, -z for 1pl with other tenses.Footnote 198 The borrowing of tense-person markers, and for that matter, tense markers in general, is extremely rare and when it occurs is indicative of intense contact. In the Balkans, we have unambiguous examples of the former in Romani borrowing from Turkish (§6.2.1.1.1) and from Greek and South Slavic (§6.2.1.1.2), a possible example involving Meglenoromanian from Macedonian which ultimately is inconclusive but interesting to explore (§6.2.1.1.3), and a partial example involving Albanian from Macedonian (§6.2.1.1.4), as well as, finally, a couple of possible examples involving tense alone, one regarding a possible, but ultimately unlikely, Albanian borrowing from Latin (§6.2.1.1.5) and the last regarding a rather compelling case involving Bulgarian borrowing from Greek (§6.2.1.1.6).
6.2.1.1.1 Romani from Turkish (and Possible Slavic Influence)
The basic Romani tense/aspect system consists of two oppositions, which produce four paradigms, as illustrated in Table 6.16.
| Imperfective | Perfective | |
|---|---|---|
| non-remote | kerel(a) = present [3sg] | kerd(j)um = preterite (aorist) [1sg] |
| remote | kerelas~kerela sine = imperfect [3sg] | kerd(j)umas~kerd(j)um sine = pluperfect [1sg] |
Romani has an opposition between a so-called long present ending in -a and a short present that does not (indicated by parentheses).Footnote 199 In many dialects in the Balkans, the long present is indicative and the short present is subjunctive (used after the dms te and the fut marker ka, etc.), although this distinction is not consistently maintained in all dialects. In fact, the neutralization tends to occur in dialects in contact with Balkan Slavic, which also lacks a morphological opposition indicative/subjunctive (cf. Cech & Heinschink 2002; Friedman 2018b, 2024).Footnote 200 Imperfects and pluperfects using invariant sine (variants hine, ine) – the 3sg preterite of ‘be’ – for all persons are typical of those Arli dialects where -s is lost in final and intervocalic position. This may be influenced by the Macedonian use of the 2sg/3sg.impf beše, which, in addition to marking the pluperfect when used with the verbal l-form, e.g., beše napravil ‘you/he had done’ can be used in any person of the past indefinite as an emphatic past, e.g., beše sum napravil ‘I have/had indeed done.’ In our glosses, we use impf and plu for those dialects with synthetic imperfects and pluperfects and rem when the marking for these is analytic.
Certain Romani dialects of eastern Bulgaria, Greece, and Crimea conjugate verbs of Turkish origin using Turkish TAM markers. The degree of realization of this phenomenon varies from a single tense – in which case it is the preterite in the suffix -DI, as in Varna Burgudži – to, in the most extreme instances, all preterites, presents, the optative, the future, negation, and even the infinitive, as in Šumen Xoraxane (Friedman 2013b).Footnote 201 Many or most Romani dialects with Turkish conjugation fit the borrowed Turkish verbs into the Romani TAM system, which is to say such verbs use the Turkish DI-past for the simple past, one of the Turkish presents for the present (see Friedman 2013b) and then Romani agglutinative and analytic markers for the remaining verbal categories. Table 6.17 gives an illustrative example from such a system from Agía Varvára (Athens), a South Vlax dialect, based on Igla 1996: 61–64.
| Nonremote | ‘do’ Rmi | ‘wait’ Trk.cnjn | remote | ‘do’ Rmi | ‘wait’ Trk.cnjn |
|---|---|---|---|---|---|
| present | kerav | beklerim | imperfect | keravas | beklerimas |
| preterite | kerdem | beklerdum | pluperfect | kerdemas | beklerdumas |
| future | ka kerav | ka belkerim | conditional | ka kerdemas | ka beklerdumas |
It is worth noting that in the case of Agía Varvára Romani, as in a number of the other dialects, speakers no longer know Turkish and are not in contact with it. The presence of Turkish conjugation in these dialects cannot, therefore, be considered a codeswitch.
Table 6.18 gives a synopsis of the Turkish verb forms that occur in a variety of Romani dialects in the Balkans. As can be seen, there is an implicational hierarchy. All dialects with Turkish conjugation have the past tense in -DI, and most have one or both of the possible Turkish present tenses (progressive in -y(or) and gnomic in -r; see Göksel & Kerslake 2005: 126–143 for discussion of various verbal categories in standard Turkish). Those with additional categories always have both present tenses.
| VB | SM | Sp | AV | FA | VD | SK | VK | SN | Fu | Ko | Ks& ŠX | VG | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| -DI | + | + | + | + | + | + | + | + | + | + | + | + | + |
| -r | - | - | + | + | + | + | + | + | + | + | + | + | + |
| -y[or] | - | + | - | - | +? | + | + | + | + | - | + | + | + |
| idi | - | - | - | + | + | + | + | + | + | + | |||
| opt | - | - | - | - | - | - | - | te+ | te+ | te+ | ø+ | (te)+ | ø+ |
| -mIş | (pt) | - | - | - | - | - | (pt) | - | (pt) | + | - | + | + |
| fut | - | - | - | - | - | - | - | - | - | k+opt | + | + | + |
| neg | - | - | - | - | - | - | - | - | - | - | + | + | + |
| inf | - | - | - | - | - | - | - | - | - | - | - | + | + |
Notes
• (pt) indicates that the perfect participle in -mIş occurs, but not the nonconfirmative finite -mIş (§6.2.5).
• For the optative, there are two types of adaptation: following the native dms te (optional if in parentheses) or without the dms (indicated by ø).
• fut = Trk fut in -AcAK, except Futadži = Rmi fut marker + optative.
Of particular interest here is the influence of Turkish on native Romani morphology in some of the dialects with Turkish conjugation. Turkish has a 1pl marker -[Im]Iz and a 2pl marker -(s)InIz. There is also a 1pl marker -k which is limited to the DI-past and the sA-conditional. In the relevant dialects, the unmarked 2pl marker -Iz is copied (with final devoicing and a generalization in vowel harmony) onto the native Romani 2pl preterite marker (-an in these dialects), and, in some cases, the Turkish -Iz is extended to the native 1pl preterite marker -am. The evidence for this extension being a Romani innovation is the fact that the Turkish dialects with which these Romani dialects are in contact have the 1pl ending -k in the preterite (Elšík & Matras 2006: 135–136; Dallı 1976: 119–124), and this is the 1pl ending used for the Turkish conjugations in these Romani dialects. Thus, for example, Turkish yaz- ‘write’ conjugates in the simple preterite (DI-past, 1sg-3pl) yazdım, yazdın, yazdı, yazdık, yazdınız, yazdılar. These same endings (ceteris paribus) are used for this verb in those Romani dialects with Turkish conjugations. The native Romani perfective preterite is formed from the root plus a participial formant that surfaces most often with a stem-final d or l (which is jotated in some dialects) plus person markers, e.g., the preterite of the native root ker- ‘do’ conjugates in many Balkan dialects (1sg–3pl): kerdum, kerdan kerda, kerdam, kerden, kerde. In the dialects noted in Table 6.19, however, the Turkish bound 2pl morpheme -InIz has been reanalyzed as -Iz, applied to the native 2pl preterite and then generalized, in some cases, to the 1pl preterite.Footnote 203
6.2.1.1.2 Romani from Greek and Bosnian-Croatian-Montenegrin-Serbian (with Slovene)
The 3sg.prs marker -i was copied into Early Romani from Medieval Greek for use with borrowed verbs, i.e., verbs marked with a special stem-formant indicating they were borrowed (i.e., xenoclitic verbs; Elšík & Matras 2006: 96–97, 326–330; see Boretzky & Igla 2004: Map 139; Boretzky et al. 2008: Map 88). In most dialects, this -i was replaced by the native marker -(e)l, but in some it survives and has even spread, e.g., in Gilan (Alb Gjilan, BCMS Gnjilane) Arli and some Arli dialects in North Macedonia and southern Serbia, where 3sg.prs -i is required for xenoclitic verbs and is facultative for polysyllabic oikoclitic verbs (Elšík & Matras 2006: 330; Boretzky & Igla 2004: Map 105; Boretzky et al. 2008: Map 61).
In BCMS (and Slovene), the 1pl present maker is -mo and, for BCMS, the 1pl aorist/imperfect marker is -smo. Some Romani dialects (or variants within a dialectal system) in contact with BCMS, and, in the case of Dolenjska Romani (also called Gopti, etc., see Cech & Heinschink 2001), also Slovene, have copied this marker – or part of it – into their person-marking systems (Cech & Heinschink 2001; Elšík & Matras 2006: 134–135).Footnote 205 Arli dialects in southern central Serbia (Jagodina, Kruševac) have been influenced by or have copied the 1pl marker -mo from BCMS, e.g., amen sijamo ‘we are,’ džanamo ‘we know,’ ka džamo ‘we will go,’ kerdžamo ‘we did/have done’ (Cech et al. 2009: 14, 24, 26). In the examples of the aorist and the present of ‘be,’ the native Romani ending is -am (in some Arli dialects 1pl aorist -em; Friedman 2013b), so only the distinctive -o of the Serbian 1pl is added. In the case of the present, however, the native ending has various shapes – -asa, -aa/-as, -a (see Friedman 2018e) – and so here we have the replacement of one morpheme by another.
6.2.1.1.3 Meglenoromanian from Macedonian
A frequently cited example of the diffusion of inflectional affixes via language contact uses data from Macedonian and Meglenoromanian that has its source in Capidan 1925a: 159–161, cited in Sandfeld 1930: 130 and Weinreich 1968: 31–32, then Heath 1984: 370, also Thomason 2001: 77, who is cited by Myers-Scotton 2002: 92. As presented in Weinreich, the first and second singular present tense markers (-um or -ăm and -iš in Weinreich, -m and -š in Thomason) were borrowed from Macedonian.Footnote 206 Both Weinreich and Thomason give the same example: aflum, afliš from aflu, afli ‘find.1sg/2sg.’ The matter has been studied more recently by Atanasov 2002, who, however, omits some of the data from Capidan. Finally, the Meglenoromanian facts themselves are not as simple as presented in Weinreich or subsequent citations. In fact, the phenomenon also has potential language-internal explanations.
There has been paradigm merger in the Macedonian dialects of the Meglen, and all verbs have the same set of stem vowels, e.g., sakum-sakiš-sake, berum-beriš-bere, nosum-nosiš-nose (‘want,’ ‘gather,’ ‘carry’ 1sg-2sg-3sg), see Ivanov 1932: 86–88; Peev 1979: 85–87; Bojkovska 2006: 90–92 for details and discussion).Footnote 207 So this is the model that was presumably available to Meglenoromanian, and in this model -um marks first person present and -iš marks second.
A careful examination of the Meglenoromanian dialectal situation, however, reveals possible internal sources for the markers, and, moreover, that this phenomenon only occurs in four of the ten Meglenoromanian villages remaining after the Balkan Wars – Nănti (Mac Noti etc., Grk Nótia), Oshinj (Mac Ošin, Grk Arkhángelos), Lundzinj (Lungonci, etc., Grk Lankadia), and Cupă (Mac Kupa, Grk Koúpa) – but not in Birislav (Mac Borislav[ci], Grk Períkle), which is located on the road from Nănti to Oshinj and Lundzinj, nor in Tsărnarekă (Mac Crna Reka, Grk Kárpi), which is the most heavily Slavicized village in terms of dialect.Footnote 208 This last fact is especially difficult for the Slavic contact argument, since Tsărnarekă is the only Meglenoromanian village where the dialect has borrowed the Macedonian gerund (verbal adverb) marker -ájkji, e.g., nirdzeájḱi ‘going’ (Caragiu-Marioțeanu et al. 1977: 209; Atanasov 1990: 2012).Footnote 209 Thus, if the person markers in question were from Slavic, one would expect to encounter them also in the only village that has unambiguously copied a bound Slavic grammatical morpheme into its verbal system. Aside from the geographic limitation, the affixes are also paradigmatically limited to a few 1st conjugation verbs with stems ending in a consonant cluster whose second element is almost always a liquid, /n/, or /k/, which would produce an unacceptable final cluster if the unaccented final high vowel were dropped as often happens in Balkan Romance. The Lundzinj form mučim ‘I bite’ – apparently a back-formation from mučiš ‘you bite’ – and the Nănti and Oshinj aflăm/aflăš (older aflu/afli) indicate that the markers in question are -m, -š. We can then note that -m, and -š are native 1sg and 2sg markers elsewhere in Meglenoromanian conjugations, e.g., IV conjugation verbs have 2sg.prs -sh, 1sg.impf is -ám or -eám, and there are also native săm ‘I am,’ ăm ‘I have’ (see Friedman 2009a for additional details). Although Nănti was by far the largest Meglenoromanian village in the region (with a small Romani minority), it was also Muslim, whereas Oshinj, Lundzinj, and Kupă are all Meglenoromanian Christian. Thus, while Slavic cannot be ruled out as one of the possible sources of the Meglenoromanian innovation (the nearest village to Nănti was Tušim (Grk Kalógeros), which was entirely Macedonian until the Greek Civil War), and the innovation might have started in Nănti and then spread unevenly outward, the details of its distribution and the existence of internal sources (which nonnative speakers in Nănti might have seen as analogous to the Slavic) do not permit us to identify with certainty a Macedonian source for this feature in the Meglenoromanian villages in question. The absence of the phenomenon from the one village that unquestionably borrows Slavic bound morphemes adds to the questionability of this example, which is therefore more a cautionary one about difficulties in teasing out diffusion (copying) versus transmission (analogy, drift). The example involving Romani and Turkish given in §6.2.1.1.1 above is much clearer, and demonstrates, as does the Tsărnarekă gerund and pace, e.g., Labov 2007: 348–349, that bound morphology can indeed be unambiguously transferred across languages.
6.2.1.1.4 Albanian from Macedonian
Almost all Albanian dialects, like the standard language, distinguish the present indicative from the present subjunctive in the second and, slightly less often, third persons of most verbs.Footnote 210 In most dialects, the two persons are marked by -n or -Ø in the indicative and by -sh and -jë, respectively, in the subjunctive.Footnote 211 In the town dialect of Debar (Alb Dibra) as well as in Tërbaç across the border in Albania, in the Dibra municipality Gjinari 2007:328-338), however, under influence from Macedonian, which, like all of Balkan Slavic, lacks a formal distinction between indicative and subjunctive, the indicative and subjunctive have fallen together, with -sh marking the 2sg person and -n marking the 3rd.Footnote 212 For the 2sg, the identity of the Macedonian and Albanian forms are independent results of changes in the Indo-European common source (*-si), but at the same time the spread of Albanian -sh is clearly influenced by its identity with the Macedonian marker. It is the restriction of this phenomenon to precisely a small region of known, intense, Albanian–Macedonian bilingualism that assures this as a contact-induced change. We can note too that loss of the formal indicative/subjunctive distinction occurs in some Albanian dialects in Kosovo (Rusakov 2010).
6.2.1.1.5 Common Albanian and Latin
Sh. Demiraj 1985: 793–795 summarizes the nineteenth-century arguments suggesting that the intervocalic -v- in the 1sg/2sg aorist of vocalic stems in Albanian, e.g., lan ‘wash’ ~ 1sg.aor lava, 2sg.aor lave; këndon ‘sing’ 1sg.aor këndova, 2sg.aor këndove are modeled on the Latin imperfect of the type cantabam, etc. or the Italian cantavo etc. He concludes, however, along with Pedersen and others, that the 1sg/2sg.aor -v- here is the result of a hiatus. Intervocalic -b- in Latin loans would have disappeared (cf. kal ‘horse’ from caballus), and the Latin-based argument does not account for the rest of the conjugation, especially the 3sg.aor of the type lau, këndoi, which are normal Albanian diphthongs versus sequences of the type oa, oe, aa, ae, which are not. Moreover, the respective stress patterns of the two languages do not correspond. Rather, the most likely scenario is that the Albanian aorist reflects desinences that are cognate with, but not resultant from, the Latin perfect desinences. They are, therefore, involved instead with a merger of synthetic perfect and aorist in Albanoid (to use Hamp’s 1994a terminology). This loss of the synthetic perfect is in turn a part of the sharpening of the aorist versus present-and-imperfect aspectual and thematic distinction. The loss of the synthetic perfect occurred throughout the Indo-European languages, but independently in each of the branches. We discuss this matter here simply because it is a proposal that has come up in the literature, even if it is not to be maintained.
6.2.1.1.6 Bulgarian and the Greek Future Marker
As noted in §6.2.4.1.3, regarding the future in Balkan Slavic, there is one future marker that does not show a straightforward development from an earlier analytic formation, at least not directly. Conev 1937: 290 notes the Bulgarian dialectal future marker za which is not derivable from an earlier sequence of šte ‘wants’ with the dms da (thus unlike other variants – see §7.4.1.2.2.1). Based on its phonic similarity to Grk θα (and presumably the absence of [θ] from Bulgarian phonology) and its restriction to southern dialects of Bulgarian in contact with Greek, Conev argues that za represents a borrowing from Greek of the marker θα. The dialect geography makes this a relatively compelling case of the borrowing of a grammatical morpheme. Thus, za is only indirectly from a reduction of a once-analytic future, i.e., on the Greek side, and is not a reduction within Bulgarian itself of a native construction.
6.2.1.2 Absolute/Relative TenseFootnote 213
The phenomenon known as sequence of tenses in normative and descriptive grammar was present in Ancient Greek and Latin. At issue are sentences of the type He told me he was angry, where what he said was “I am angry.” It is found in Albanian and survives in the extra-Balkan Romance languages, and to some extent in Aromanian. Also, according to Luria 1930: 197, it is generally preserved in the Judezmo of Bitola, and occurs as well in some Macedonian dialects in Albania (Makartsev 2018: 220; Steinke & Ylli 2007: 366). Moreover, there is no evidence of it in Slavic beginning with the earliest attestations.Footnote 214 What this means is that in Balkan Slavic, Modern Greek, Romanian, and Romani, the report would (normally) be rendered in the present tense – see (6.71a–e) – whereas in Albanian and Judezmo, at least some Aromanian, and some Macedonian, there would be marking for relative tense, and the report would be shifted to agree with the past tense of the main verb; see (6.71f–g) – in (6.71h), we note the same tense agreement phenomenon but with a main verb of ability:
a.
Mi-a spus că e supărat (Rmn, Zafiu 2013a: 63–64) me.dat-has said that is angry b.
Mi reče deka e lut (Mac) me.dat said that is angry c.
Kaza mi če go e jad (Blg) said me.dat that him.acc is angry d.
Μου είπε ότι είναι θυμωμένος (Grk) me.gen say.pst.3sg comp is angry e.
Phengja mange kaj si le xoli (Rmi) said me.dat that is him.acc angry f.
Më tha se ishte i zemëruar (Alb) me.dat said that was PC angry ‘He told me that he was [=is] angry’ g.
dijo ke estaba en kasa (Jud, Symeonidis 2002: 118) say.3sg.aor that be.3sg.impf in house ‘he said that he was [=is] at home’ h.
Cu vruta nu puteam s’ aduneam (Aro, Sandfeld 1930: 117) with beloved not could.1sg dms met.1sg ‘With [my] beloved I could not meet’
In those languages that do not have so-called sequence of tenses, the use of a past tense form subordinated to a present tense has various pragmatic implications. Thus, for example, a past tense could imply that he was no longer angry, i.e., that he had been angry.Footnote 215 From a Balkan linguistic point of view, Greek probably lost sequence of tenses by the Koinē period and Balkan Romance appears to have lost the feature in the presence of Slavic, which entered the Balkans without it at the time of the last Balkan Latin inscriptions. Given the evidence of Sanskrit, Romani probably entered the Balkans without it, while in the case of Albanian our historical record is not deep enough to say what the pre-modern situation was. In the case of Judezmo, similar phenomena in French and Italian point to a possible non-Balkan Romance preservation. In view of the historical record and the situation in the extra-Balkan Romance languages, it seems fair to conclude that the lack of sequence of tenses is at least a partial Balkanism, spurred on by Slavic in the case of Balkan Romance, although, pace Asenova 2002: 260, it is found in Albanian as well as Macedonian dialects in Albania (Makartsev & Wahlströhm 2018).Footnote 216
We can also note here the phenomenon of transposition of tenses, e.g., the use of an aorist or perfect with future meaning, e.g., in the right context, the ostensible past tense forms ‘I came; I arrived,’ Alb Erdha!, Grk Έφτασα!, Blg Dojdoh!, Mac Dojdov!, Rmn Am venit!, Rmi A[v]iljum, Trk Geldim ‘I will come [right now].’Footnote 217 In this construction, in Romanian, the (etymological) perfect is generally used (cf. §6.2.1.3 below), whereas in Albanian the aorist is more common. In Balkan Slavic the aorist, which is confirmative (see §6.2.5.1 below) is more expressive than the perfect (cf. Asenova 2002: 261–262), and in Greek the perfect is not used.Footnote 218
6.2.1.3 Simplex versus Analytic Pasts
The opposition between a synthetic preterite and an analytic perfect – for which the synthetic preterite is always the older form and the analytic perfect is an innovation of sometimes greater sometimes lesser antiquity (see §6.2.3) – is characteristic of most of the Balkans. Balkan Romance in this respect reflects developments that parallel those of Balkan Slavic. The possibility that these parallels involve convergent causation cannot be ruled out, but neither can it be proved. Still, the similarities in areal distribution are striking. In its grossest terms, at issue is a split between conservative systems with a simple preterite/analytic perfect opposition and those innovating ones where the analytic perfect has become the unmarked past. Basically, Greek, Albanian, and Balkan Slavic, including Torlak BCMS, have robust systems, as do Aromanian and Meglenoromanian. Romanian and non-Balkan BCMS have gone furthest in eliminating or restricting simplex pasts at the expense of the perfect, albeit with an important difference. In Romanian, the aorist is eliminated in favor of the perfect but the imperfect is preserved, whereas in BCMS both aorist and imperfect are eliminated in favor of the inherited perfect, with imperfective aspect taking over the imperfect’s functions (see §6.2.2). Each of these tendencies is found in non-Balkan Romance and Slavic, respectively. For Romani there is no aorist/perfect opposition (see §6.2.1.1.1) except as a contact-induced innovation in a few dialects (see below and §6.2.3.4).Footnote 219 The analytic imperfect in some Arli dialects (see Table 6.16 in §6.2.1.1.1) is an internal development motived by the loss of final /s/ and following the inherited tendency to using encliticized verb forms. Like Spanish, Judezmo has retained the simple past/perfect opposition, although there is a tendency to blur the opposition between them (Gabinskij 1992: 96; §6.2.3.5).
Georgescu 1958 gives a detailed account of the distributions for Balkan Romance. Oltenia, most of Muntenia, and the Apuseni mountains – as well as Aromanian and Meglenoromanian – are the most conservative. Dobrudja, southern Moldavia, most of the Banat and adjacent parts of Crișana and Maramureș are transitional (preservation of aorist only in 3rd person), while Bucovina, northern Moldavia, northern and central Transylvania, and adjacent parts of Crișana and Maramureș are the most innovating, with the new perfect completely replacing the aorist. Southwestern Banat, where there is still a Serbian minority, is a second region where the aorist has been completely replaced by the perfect. Georgescu’s data are concerned with morphological preservation rather than morphosyntactic or semantic dominance. The picture thus differs from that given by Asenova 2002: 273, which does not differentiate between those areas of greater and lesser preservation of the simple preterite. To be sure, Asenova is attempting to portray relative frequency or dominance. Nonetheless, the areas of stronger morphological preservation arguably correspond to greater usage. According to Zafiu 2013a: 58–59, the simple past in standard Romanian is used only for literary narrative (mainly third person) and as a recent past in southern regional varieties, especially the southeast (Oltenia, where it is emblematic for the region); it is completely obsolete in northern varieties. She compares the situation with Italian, where the simple past is obsolete in the north and flourishing in the south (even more than Romanian). Another example is Banat Bulgarian, especially in Serbia, where the synthetic pasts have been supplanted by the old perfect, just as in the Serbian dialects of the same (northern) region (Nomachi & Browne 2019).
The transformation of perfects into unmarked pasts that replace the simple preterite is a phenomenon that has occurred independently in a number of places in Europe. At the same time, the preservation of simple preterites is more characteristic of European peripheries, except in the Balkans, where it is arguably a central feature, in part owing to the rise of new verbal categories (evidentiality/status). In Western Europe, the replacement of the aorist with the perfect appears to have originated in France and moved down to northern Italy and across to Germany. Peripheries such as the British Isles, Scandinavia, the Iberian peninsula, the Netherlands, as well as the Baltics and Sorbian are more conservative. With the exception of Sorbian and Balkan Slavic, a similar replacement of the simple preterite by an analytical perfect took place throughout Slavic, albeit quite independently, i.e., after the break-up of Common Slavic. The use of the perfect in consecutive narrative passages can be taken as diagnostic of its passage to an unmarked past. Narrative perfects occur in Albanian (Friedman 1983; Boretzky 1966), although synthetic pasts are still robust. The situation in Balkan Romance is noted above. In Balkan Slavic, the old perfect has taken on evidential meanings in Macedonian and Bulgarian but not necessarily in Torlak BCMS (cf. Friedman 2014b: 111–112, also Samilov 1957) and is thus narrative but limited in other ways. This situation resembles that in Turkish, and in the case of Turkish the system was already in place at the time of the earliest (eighth century) inscriptions (Tekin 1968). For Slavic in the Balkans, the robustness of the aorist and imperfect declines as one moves north and west from North Macedonia and Bulgaria. Although still robust in the Torlak dialects, in the standard BCMS languages, the aorist and imperfect are basically literary narrative devices not unlike the situation in some of the other standard languages of Europe, even if the evolution itself is independent. Colloquial BCMS usage grows more robust as one moves south and east (including, here southern Montenegro, but also Bosnia). The imperfect is weaker than the aorist (unlike the situation in, e.g., French), and there is a language ideological aspect to the situation, namely the usages are discouraged by Croatian normativists as opposed to Serbian (Friedman 2003c). Moreover, use of the synthetic pasts is also associated with Bosnia (VAF field notes 1974). We can also note here that in the Banat Bulgarian of Serbia, the synthetic pasts have been lost (Nomachi & Browne 2019). At the extreme southwest of Macedonian, i.e., the Kostur-Korča dialects, the new perfect in ‘have’ has completely replaced the old perfect in ‘be,’ except in a few nonconfirmative uses of the latter. In this, the Kostur-Korča Macedonian dialects have come to resemble most closely Albanian, including those dialects with which it has been in contact, which has a marked nonconfirmative (see §6.2.5).
Greek and Macedonian, as well as Parakálamos Romani and some other dialects (Matras 2004), have resultative perfect paradigms and simple preterites.Footnote 220 Most dialects of Romani use the preterite for both simple preterite and perfect meaning. Some Romani dialects with Turkish conjugation, however, have the option of deploying the additional past tenses of Turkish with Turkish verbs (see Table 6.18, §6.2.1.1.1). In Judezmo, some dialects have completely lost the older analytic perfect in ‘have’ (aver) and replaced it with a newer one in ‘hold’ (tener). Such usages, however, are also attested in Asturian, Mirandese, and Portuguese, with the former having more or less eliminated the simple preterite, and, as noted above, Judezmo shows a tendency to blur the distinctions between the two types of pasts (Gabinskij 1992: 96).
We can also note here a difference among the Balkan languages in the use of aorist versus perfect in formulaic expressions such as ‘welcome’-‘well met’ (see §4.3.10.2.2). Most of the Balkan languages use an aorist, but Romanian and Bulgarian use a perfect.Footnote 221
6.2.2 Aspect: Aktionsart, Superordinate, SubordinateFootnote 222
In examining Balkan aspect we can distinguish three types (cf. Aronson 1981): aktionsart, superordinate aspect, and subordinate aspect. In the context of Slavic, aktionsart has an effect on both the aspectual and the lexical meaning of a verb and is usually marked by prefixation, although suffixation (including stem-vowel alternation, e.g., Macedonian frla/frli ‘throw imfv/pfv’) is also occasionally utilized. Aktionsart generally has the effect of converting an imperfective or bi-aspectual to a perfective.Footnote 223 Thus, in Balkan Slavic, aktionsart straddles lexical and grammatical derivation. Superordinate aspect refers to the opposition labeled perfective/imperfective in Slavic, Greek, and Romani. In structural terms, it is superordinate to the tense opposition past/nonpast or, in the case of Romani, to another aspectual opposition (remote/nonremote, see Table 6.16 above). In the Balkan context, subordinate aspect refers to an opposition that occurs only in the past tense and is thus subordinate to tense in a structural hierarchy: this is the aorist/imperfect opposition within the past tense.Footnote 224 Subordinate aspect is often involved in debates over relative tense, but in the Balkans, those languages with both oppositions clearly treat them as aspectual. In Greek and Romani, aorist and imperfect align unambiguously with perfective and imperfective, i.e., the superordinate aspectual distinctions are realized in the past as aorist/imperfect. In Albanian, the opposition always involves past tense, and it can be projected into forms marked for resultativity and taxis owing to the aorist/imperfect opposition in auxiliaries ‘be’ and ‘have.’ In Balkan Romance, the aorist-perfect/imperfect opposition is described in terms of perfective and imperfective aspect (Zafiu 2013a: 55).
For Balkan Slavic, the fact that aorists and imperfects can both be either perfective or imperfective leads to a debate concerning whether the imperfect is marked for temporal coordination or aspectual duration. It is sometimes claimed that the imperfect must always coordinate with some other past event. While the behavior of the Balkan Slavic imperfect resembles a mapping of the present onto the past, the crucial fact that an imperfect can be followed by an aorist that is temporally subsequent to it demonstrates that at issue is an aspectual opposition involving duration, as in example (6.72) from Friedman 2014b: 26 and example (6.73) cited in Friedman 1993c:
(6.72)
Rabotnički igraše prvoligaški vo Kosovska Mitrovica i Workers play.3sg.imfv.impf big.league in K. M. and ottamu se vrati so dva boda. thence refl return.pfv.aor with two points ‘[The soccer team] Workers played big league ball in Kosovska Mitrovica and returned from there with two points.’
(6.73)
Prvin rabotev vo drvodelskiot artel, potoa at.first work.1sg.ipfv.impf in carpenter’s.def workshop then preminav vo fabrika, naučiv za bravar. (Šolohov 1970: 7) move.1sg.pfv.aor in factory learn.1sg.pfv.aor for locksmith ‘At first I was working in the carpentry shop, then I moved to a factory and learned to be a locksmith.’
Aside from the importation of Turkish subordinate aspectual distinctions into some Romani dialects with Turkish conjugation mentioned in §6.2.1.1.1 above, and the creation of an aorist/imperfect distinction in some Aromanian compound pasts (§6.2.2.1.1 below), subordinate aspectual distinctions are less likely to affect another language in a contact situation (but see §6.2.2.1 on Macedonian), while aktionsart, owing to its quasi-lexical nature, is more likely to do so. It has even been suggested (e.g., Asenova, 2002: 264–269) that the relative conservatism in the preservation of subordinate aspectual distinctions is a Balkanism, i.e., a linguistic feature shared and selected on the basis of multilingualism. It is certainly the case that for both Balkan Romance and Balkan Slavic as one moves north and east out of the main Balkan contact zone, the use of the synthetic pasts declines at the expense of the old perfect (cf. §6.2.1.3). On the other hand, insofar as we demand, as do Neogrammarians past and present (Eric Hamp, p.c.), that only the shared innovations be counted as diagnostic, such a shared archaism would not be treated as contact induced. Nonetheless, there are shared specificities at all aspectual levels in the Balkans that can reasonably be attributed to the influence of multilingualism.
6.2.2.1 Aorist/Imperfect
As indicated above, the fact of the preservation of an aorist/imperfect subordinate aspectual opposition is a congruent retention in the various Balkan languages and therefore not in itself a Balkanism. Nonetheless, there are some features that deserve comment here. One is that west Macedonian dialects are distinguished from those in the east as well as from Bulgarian by the tendency to make subordinate and superordinate aspect congruent. This tendency can be seen in light of contact with the local non-Slavic Balkan languages and is an on-going process. Aorists in these dialects are formed exclusively or almost exclusively from perfective verbs, and imperfective aorists are basically obsolete.Footnote 225 For imperfects, the perfective imperfect cannot occur independently, but only after the same modal particles to which the perfective present is limited. Thus, there are contexts in which Bulgarian requires an imperfective aorist where the choices in Macedonian are either perfective aorist or imperfective imperfect, as in example (6.74) where the combination of short time duration and limitation requires an imperfective aorist in Bulgarian (trèse) but an imperfective imperfect in Macedonian:
(6.74) a.
Zemjata se treseše tri sekundi earth.def refl shake.3sg.impf three seconds ‘The earth shook for three seconds’ (cf. Friedman 2014b: 27; cf. Dejanova 1966: 58)
The Greek equivalent of this sentence would normally employ the aorist, i.e., a perfective. However, native speakers were comfortable with the imperfect if the duration of the event were being stressed, e.g., ‘three whole seconds’ or with a longer time period, e.g., ‘one minute’ (which is a long time for an earthquake):
b.
(Ταρα)κουνήθηκε το έδαφος για 3 δευτερόλεπτα shake.3sg.aor the.nom.sg ground for 3 seconds c.
κουνηθήκαμε για 3 δευτερόλεπτα shake.1pl.aor for 3 seconds d.
κουνιόμασταν ένα λεπτό (or for other longer periods) shake.1pl.impf one.sg minute e.
η γη σείστηκε για 3 δευτερόλεπτα [only this; i.e., *σειότανε (impf)] the.nom.sg earth quake.3sg.aor for 3 seconds f.
η γη σειότανε για 3 (ολόκληρα) δευτερόλεπτα the.nom.sg earth quake.3sg.impf for 3 whole.pl.acc seconds
The situation is the same in Albanian. The short time limitation precludes the use of the imperfect, but for a longer time period the imperfect, and even the progressive imperfect, can be used to stress the duration.
g.
Dheu u dridh për tri sekonda earth.def refl shook for three seconds h.
Dheu [po] dridhej me orë të tëra earth.def prog shake.impf.mdp with hour PC entire
This simplification in western Macedonian leads to overall greater congruence with the non-Slavic verbal systems, even though the specifics of aspectual assignment may differ. The perfective aorist corresponds to the aorist in languages without superordinate aspect and also to the Greek, where aorists must be perfective (see also §6.2.2.2.3 below).Footnote 226
Albanian and Balkan Romance have an imperfect/aorist opposition for ‘be’ and ‘have.’ In principle, Balkan Slavic has an imperfective aorist of ‘be’ (3sg Mac bi Blg be), but in practice it is limited to a few expressions. The aorist of ‘have’ would be homonymous with the imperfect except the 2sg/3sg, which look like presents, but in fact the aorist does not occur. In Greek, ‘be’ and ‘have’ are imperfectives.Footnote 227 Albanian is unique in using both the aorist and imperfect of auxiliaries in forming the perfect and pluperfect, but this pattern was replicated by Gorna Belica Frasheriote Aromanian under the influence of Albanian (see §6.2.2.1.1 below).
6.2.2.1.1 Subordinate Aspect: Aromanian from Albanian
As just indicated above in §6.2.2.1, the Frasheriote dialect of Gorna Belica (Aro Bela di Suprã), North Macedonia, whose speakers arrived from Myzeqe in Albania about a century ago, have calqued the Albanian subordinate aspectual distinctions into their analytic past tenses. As Markovikj 2007 observes, this continues to serve them in translating the Macedonian superordinate aspectual distinction as well, e.g., Frasheriote Aromanian ave lukrată/avu lukrată ‘he has [impf/aor] worked,’ based on Albanian kish punuar/pat punuar ‘idem,’ now corresponds to Macedonian ima raboteno/ima sraboteno ‘he has worked’ [ipfv/pfv].Footnote 228 See Table 6.31 in §6.2.5.6 for the relevant indicative and admirative forms of Albanian and Gorna Belica Aromanian. We have in the example here an instance of contact-induced change involving subordinate aspect corresponding to superordinate aspect. In this instance, the forms were available to be utilized, but only as a result of the Albanian model was this utilization realized in the given Aromanian dialect.Footnote 229
6.2.2.2 Perfective/ImperfectiveFootnote 230
The superordinate aspectual opposition perfective/imperfective is found in Balkan Slavic, Greek, and Romani. In each of the relevant languages, however, the morphology and temporal limitations differ. In the case of Slavic, both Balkan and non-Balkan, superordinate aspect is inherent in the lexical verb, and some lexical verbs are bi-aspectual. The opposition was still in a formative period at the time of the break-up of Common Slavic, and the details of the opposition differ from language to language, even those so closely related as Macedonian and Bulgarian (see Friedman 1985b). Moreover, the superordinate aspectual opposition in Slavic, being inherent in the lexeme, is also involved in derivation. Prefixation, which can also involve aktionsart and change in lexical meaning, generally renders an imperfective perfective, e.g., Mac bara/pobara ‘seek [ipfv/pfv],’ drži/zadrži ‘hold, detain, delay, hold back [ipfv/pfv].’ Suffixation usually renders a perfective imperfective, e.g., Mac kaže/kažuva ‘say [pfv/ipfv],’ but certain suffixes are perfectivizing, e.g., BSl čuka/čukne ‘knock.3sg [pfv/pfv].’ In some cases, the difference is encoded only in the stem vowel, e.g., Mac frli/frla ‘throw [pfv/ipfv].’
In Greek, superordinate aspect is an opposition that is realized within each lexical verb and is expressed almost exclusively by means of suffixation within the lexical item, e.g., (using the present for ‘write’ plus ‘fut’), ipfv.fut θα γράφ-ω ‘I will be writing’ and pfv.fut θα γράψω (i.e., γράπ-σ-ω) ‘I will write.’Footnote 231 Similarly, the ipfv/pfv opposition is found in the imperative, e.g., ipfv.impv γράφε(τε) ‘Be (i.e., continue) writing.sg/pl!,’ pfv.impv γράψε/γραψτε ‘write.sg/pl!’. In the past, the imperfect tense is only imperfective and the aorist tense is only perfective, e.g., 1sg ‘write’ ipfv.impf έγραφα ‘I was writing’ ~ pfv.aor έγραψα ‘I wrote.’ The perfect and pluperfect are likewise only perfective, e.g., 1sg ‘write’ prf έχω γράψει ‘I have written ~ plu είχα γράψει ‘I had written.’
This situation differs from Balkan Slavic, where imperfects and aorists can both be perfective or imperfective, although, as noted above (§6.2.2.1), the ipfv.aor is now obsolete for most Macedonian speakers. The perfective imperfect, however, is still a vital part of all Balkan Slavic verb systems.
Unlike the situation in Greek, perfects in Slavic also have the imperfective/perfective opposition, and in some cases, also the aorist/imperfect opposition. For Macedonian, the example imam raboteno/imam srabeteno ‘I have worked [ipfv/pfv],’ cited above in §6.2.2.1.1, is illustrative. In Bulgarian, kazval e/kazal e ‘he has spoken/said [ipfv/pfv]’ is a typical example. In Torlak BCMS, too, the imperfective/perfective opposition extends into the l-past. The Common Slavic resultative participle, which is retained as such in Bulgarian and which becomes the verbal l-form in Macedonian, was formed from the aorist stem, and so the opposition aorist/imperfect did not extend into the inherited perfect.Footnote 232 This opposition, however, was extended into the l-form for most of Macedonian and Bulgarian in connection with the development of evidentiality. See §6.2.5 for additional details.
The past passive participle, which remained as such in Bulgarian and became the verbal adjective in Macedonian and the basis of its ‘have’ perfect, is normally formed using the aorist stem for verbs with the stem vowel -a in the aorist and the imperfect stem vowel for others. For a limited class of Macedonian verbs, however, it is possible to have an aorist/imperfect alternation in the verbal adjective, and therefore in the ‘have’ series of perfects and pluperfects as well. See Friedman 1993c for additional details.
The Romani situation is similar to that in Greek, but mostly less complex, since Romani – unlike Greek and Balkan Slavic – generally has no perfective/imperfective contrast within the present. Rather, as Matras 2002: 151–152 observes, Romani has a secondary aspectual opposition, remote/nonremote, and the imperfective-remote is imperfect vis-à-vis the present and the perfective-remote is pluperfect vis-à-vis the aorist, which also generally functions as the perfect. Thus, the Romani present/subjunctive opposition, to the extent that it is realized, does not represent an aspectual contrast.Footnote 233 To be sure, individual dialects copy Slavic prefixes for derivation, and the lexical items involved (e.g., kinel/pokinel ‘buy,’ del/dodel ‘give’) can be aspectually specialized. Still, these are basically lexical rather than grammatical phenomena insofar as they do not pervade the verbal system.
As noted above, Greek has extended the imperfective/perfective opposition into the future tense, since there is now a distinction between θα γράφ-ω ‘I will be writing’ and θα γράψω (i.e., γράπ-σ-ω) ‘I will write’ which was not possible in Ancient Greek with its monoaspectual (and apparently only perfective) future, e.g., γράψω ‘I will write.’ This distinction developed only with the ‘want’ based (and pan-Balkan) periphrastic future of Medieval Greek; the earlier Postclassical future periphrasis with ‘have’ did not allow for imperfective/perfective differentiation. Greek has always had a distinction between imperfective (presential) and perfective (aoristic), generally encoded by different stems (e.g., prs.ipfv in -ιζ- vs. aor.pfv in -ισ-); in the Middle Greek period there was much reshaping of the marking of the different stems (Horrocks 2010:303-316), usually by reuse (spread, with some reanalysis) of an Ancient Greek suffix. One fairly productive overt marker that arose innovatively (partly an adaptation of the earlier -νυ- present suffix) was the suffix -ν- ([n]), for imperfective aspect on many verbs, generally imperfectivizing an aorist stem. Horrocks 2010: 305 describes it as “the new principle of substituting imperfective [-n-] for aorist [-s-]”; in some instances, it involves re-shaping the imperfective part of an Ancient Greek opposition of present versus characterized (marked) aorist), e.g., λύ-ν- ‘loose’ (versus perfective (aoristic) stem λύ-σ-), corresponding to AGrk λύ- / λύ-σ-, although many presents lack the nasal, e.g., γράφ-ω ‘I write, I am writing,’ or enter into other marking schemes, as with -ιζ-/-ισ-. Still, the upshot is that imperfective aspect has a consistent overt mark in Modern Greek that was absent in earlier stages. (cf. also §7.7.2.1.3.2.3). It is worth noting that Romani borrowed the marker -ιζ- (as -iz-) as a productive loanverb adaptation marker during the Early Romani period, although different dialects have selected among this and other markers, e.g., -in-, also from Greek (Boretzky & Igla 2004: Map 139).
As also noted above, the imperfective aorist is almost extinct (its functions replaced either by perfective aorists or imperfective imperfects) in Macedonian, although it is still vital in Bulgarian. Another feature of Macedonian aspect which, while not necessarily contact related, is nonetheless leading to greater aspectual transparency and regularity is the elimination of biaspectual verbs in -ira by adding perfectivizing prefixes, with iz- being by far the most common (Markovikj 2010), e.g., organizira/izorganizira ‘organize pfv/ipfv.’ As recently as the 1970s, such prefixation was excluded from the norm and strict bi-aspectuality was prescribed for verbs in -ira. Of particular significance is the fact that 3sg present and 3sg aorist were thus homonymous. In the 1970s, the ambiguity was avoided by using the 3sg imperfect, but apparently speakers were dissatisfied with this solution and began using prefixes. The tendency, therefore, is to place more aspectual weight on transparent prefixal markers, which is consistent with contact-induced change as well as internal pressure. The fact that approximately a third of the population of North Macedonia has a home language other than Macedonian might or might not be a contributing factor, but it certainly has not hindered the development.Footnote 234
One other aspect-related commonality among the Balkan languages relates to the borrowing of aorist (perfective, simple preterite) morphemes for loanverb adaptation (cf. §4.2.2.2) and the frequent appearance of aoristic stems in verbal borrowings. Thus, in all the Balkan languages, the Turkish DI-past can serve as the basis for the borrowing of a verbal stem, and in some languages other than Greek, the -s- of the Greek sigmatic aorist can also be added e.g., Mac bendisa, Alb begendis ‘be pleasing, like’ (see §4.2.2.2 for other examples). Moreover, in verbal loans directly from Greek, s-aorist stems are normally found, e.g., Pomak diakonisovam ‘I serve,’ from Greek διακονώ (aorist stem διακονησ-; see Joseph 1987b for other examples). In the case of Albanian, there is a native sigmatic conjugation (which is also used for adapting Slavic loanverbs).
Overall, it can be said that perfectivity, often marked by preverbs, is borrowed, while imperfectivity tends to be calqued (although, as noted above, the Greek imperfectivizer -ν- (-n-) becomes a simple loanverb adaptor in some Romani dialects).
6.2.2.2.1 Perfective/Imperfective: Meglenoromanian from Macedonian
Meglenoromanian has a Slavic-like perfective/imperfective opposition, e.g., durmíri ‘sleep’ ~ zădurmíri ‘fall asleep’ (Atanasov 2002: 226–227; cf. Capidan 1925a: 196–202; Saramandu & Nevaci 2020: 163–164). Meglenoromanian also uses Macedonian aspect/aktionsart preverbs to mark iterativity, e.g., native turnári ‘return’ ~ priturnári ‘return repeatedly,’ and Slavic văvíri ‘go/set out,’ ~ zăvăvíri ‘go/set out repeatedly.’ Macedonian also has vrvi/ zavrvi, but the opposition here is imperfective/perfective. For Macedonian, to imperfectivize zavrvi, an imperfectivizing suffix, -uva, would be expected: zavrvuva. Such a form does not occur in the standard language, but as Humphries 1997 has shown, -uva is even more highly productive in colloquial Macedonian than in the prescribed norm for deriving imperfectives from perfectives, and this is especially true in the Meglen dialect of Macedonian (Bojkovska 2006: 86). Even in Standard Macedonian, the regularization of imperfectivization is much more productive than the use of the same or related suffixes in any other Slavic language. When Meglenoromanian copies a Macedonian semelfactive perfective marked with the infixal perfective marker -n, e.g., čukníri ‘knock’ (Mac čuka/čukne ‘knock.pfv/ipfv’), the Meglenoromanian prefixed iterative uses the Macedonian imperfective base: zăčukăjíri ‘keep knocking,’ cf. Macedonian perfectivized začuka ‘begin to knock,’ which is imperfectivized as začukuva. Elsewhere Aromanian copies the Macedonian imperfective/semelfactive pair, e.g., lăpăjíri ‘swallow,’ lăpníri ‘gulp suddenly/all at once; take a swallow,’ cf. Macedonian lapa, lapne ‘idem.’ It would appear that Meglenoromanian has followed the Macedonian tendency of productively deriving iteratives, but rather than generalizing -uva, Meglenoromanian has adopted the strategy of extending the productivity of prefixation (see also Saramandu & Nevaci 2014b).
6.2.2.2.2 Aktionsart~Perfective/Imperfective: Romani from Slavic
Romani dialects are in contact with a wide variety of languages with various types of aspectual marking (cf. Matras 2002: 117–164; Elšík & Matras 2006: 188–202). Those dialects in contact with Slavic both inside and outside the Balkans (for North Russia, see Rusakov 2004; also Matras & Elšík 2006: 193 for Slovak as well as non-Slavic Hungarian). Thus, for example, Arli Romani in North Macedonia kinel ‘purchase [in general]’ > pokinel ‘buy [take possession of],’ lel ‘take, get, begin (in general)’ > polel ‘grab, take hold of, seize,’ džakerdo ‘awaited’ > dodžakerdilo ‘expected, anticipated.’ Unlike the Meglenoromanian phenomenon, which pervades the system, the influence of Slavic on Romani is more like that of lexicalized aspect. See also Igla 1998 for a detailed account of Bulgarian Romani aspect.
6.2.2.2.3 Aspect Neutralization: Macedonian from Aromanian (with Albanian and Greek)
Markovikj 2007 discusses the use of verbal nouns in Ohrid Macedonian on the model of Aromanian verbal nouns. The Ohrid Macedonian usage reflects the Aromanian adaptation of Macedonian’s superordinate aspectual distinction, since Aromanian lacks the Macedonian distinction, and Macedonian verbal nouns are always derived from imperfective verbs. Thus we have an Aromanian adaptation spreading into Macedonian, as is the case with other phenomena (see, e.g., §6.2.3.2.1 below).
In Frasheriote Ohrid Aromanian (and also Kruševo – Gołąb 1984a: 106), the gerund (the old present participle in -inda/-ânda) has disappeared completely. The corresponding Macedonian verbal adverb (standard -é/ájkji, with many dialectal variants), descended from the old present active participle and derived only from imperfective verbs is likewise extremely rare in Ohrid. On the basis of the situation in Romanian, Markovikj 2007 argues that the Aromanian gerund likewise had a variety of functions that differed from the purely adverbial function in Macedonian already attested in the late medieval period along with the Macedonian form’s inherited limitation to reference to the subject (see also §7.7.2.3.1.2 on subject-only coreference). Therefore, to render communication more isomorphic, speakers brought the two systems closer to one another by adopting constructions using the verbal noun (which in Aromanian is descended from the old infinitive). The result is both a reinstantiation of nonfiniteness in the local Macedonian dialect and a neutralization of the imperfective/perfective opposition. Instead, Macedonian verbal nouns derived from iteratives are preceded by edno ‘one’ to render a perfective effect by focusing on a single iteration. Consider the examples in (6.75) from Markovikj 2007: 75 (with normalized orthography here):Footnote 235
i.
Tu muntreri televizie, me akãcã somu. (Ohrid Aro) in watching television me grab.3sg.aor sleep ii.
Vo gledanje televizija zaspav. in watching television fall.asleep.1sg.aor (Mac) ‘While watching television, I fell asleep.’
i.
‘Unã shãderi, gãdzui. (Ohrid Aro) one sitting fall.1sg.aor ii.
Edno sedvenje, padnav. (Mac) one sitting fall.1sg.aor ‘When I sat down, I fell.’
In standard Macedonian, example (6.75aii) would use an imperfective gerund gledajkji ‘while watching,’ whereas in example (6.76bii), either a derived iterative imperfective gerund sednuvajkji ‘while in the act of sitting down’ or a perfective štom sednav ‘when I sat down’ would be used. The Aromanian constructions in the (i) examples use a verbal noun, muntreri and shãderi respectively, and a quantifier, e.g., ‘Unã in (6.75bi) if necessary, to limit the action, and the Ohrid Macedonian has calqued this, in part as a substrate effect. Thus, here, as in other parts of the grammar of Ohrid Macedonian, it can be argued that the Balkan Romance substrate has reinforced certain kinds of feature selection (cf. Mufwene 2008).
A similar usage can also be found in some Romani dialects in the Balkans, as in the following Skopje Džambaz (South Vlax) example (Bodnárová 2018b) with a gerund:
(6.76)
Gova vreme, ek phure sovi, mora te uštol that time one old.pl sleep.grd must dms get.up.3sg ‘In those days, when the old folks were asleep, she [the bride] had to get up’
The use of ‘one’ plus verbal noun to describe a sudden sequence of events also occurs in Albanian, as seen in examples (6.77):
a.
Një të marrë flakë dhe u bë shkrumb e hi one pc take.ptcp flame and refl become.3sg.aor cinder and ash ‘It caught fire and became [nothing but] cinders and ashes.’ (i.e., ‘Once it caught fire it was reduced to … ’; Buchholz & Fiedler 1987: 183 = Newmark et al. 1982: 101)Footnote 236 b.
Një të parë atë, sakaq, edhe nxënësit e tjerë one pc see.ptcp him.acc immediately and pupils.def.nom.pl pc other filluan të lëviznin. begin.3pl.aor dms move.3pl.impf ‘Having seen (lit., ‘one seeing’) him, immediately the other pupils as well began to move.’ (i.e., ‘Once they saw him, … ’; Newmark et al. 1982: 101)
The use of ‘one’ + verbal noun (or gerund in the Vlax case cited above) to describe a sequencing of events is specific to the region where Aromanian, Macedonian, and Albanian converge. The usage can also mean ‘the moment when, or just [do X]’ as in the following Macedonian example (6.78) (cf. Sandfeld 1930: 126, and §7.7.2.1.2):
(6.78)
Edno vleguvanje vo klinika i vednaš se dobiva one going to clinic and at.once refl gets jasna slika kolku ni e raspadnata država clear impression how to.us is ruined state ‘Just go into a clinic and right away one gets a clear impression how ruined our state is’ (lit., ‘an entering in [a] clinic and right away … ’)
Other Balkan languages do have idioms involving ‘one’ and verbal forms, but they differ in meaning and, sometimes, in choice of form, and none of them have the perfectivizing meaning of the type ‘Upon …, ’. Thus, for example, in Bulgarian, edno plus verbal noun is used to express emphasis on the action, e.g., edno piene ‘such drinking!,’Footnote 237 but the usage of the type found in Macedonian is completely unknown.Footnote 238
Also, Greek can use ‘one,’ in its feminine form, plus ‘and’ or ‘that’ plus the aorist to mean ‘since, as long as … ,’ e.g., (6.79):
(6.79)
μια και/που ήρθες one.f and/that come.pst.2sg ‘Since you have come, [let’s eat]’
At issue in the effects of language contact on tense and aspect systems is how speakers sense differences on the basis of surface phenomena. The use of ‘one’ plus verbal noun precisely in Albanian, Aromanian, and Macedonian but not in Bulgarian or Romanian points to a convergent aspectual expression that neutralizes the contrast between Slavic and non-Slavic types of aspectual systems. The Romani example is, appropriately, using ‘one’ + grd to render a kind of middle ground.
6.2.2.3 Intersection of Superordinate Aspect and Mood
Languages with the superordinate aspectual distinction (Balkan Slavic, Greek) have special restrictions on the occurrence of perfective presents and imperfects, the most common and prominent among those being combinations with the dms and the future marker. Those Balkan languages without superordinate aspect (Balkan Romance and Albanian) have morphological present subjunctives (and perfects when the present subjunctive is used as an auxiliary) that are restricted to the dms, with the occasional addition of the future marker (< fut + dms). Thus in Greek and Balkan Slavic, the perfective present form is normally found only after the dms, the future marker, and a limited number of other words that vary to some extent according to language.Footnote 239 The same restrictions apply to the perfective imperfect in Balkan Slavic.Footnote 240 Independent uses are either highly restricted or nonexistent. In Balkan Romance and Albanian, the subjunctive is likewise restricted to occurrence with the dms, sometimes also the future marker, and it occurs independently only rarely or not at all. Romani occupies a midway position. Like Greek, imperfect and aorist line up with imperfective and perfective aspect, but owing to the other aspectual opposition of remote/nonremote, there is no perfective present (or perfective imperfect). Many Romani dialects in the Balkans have a morphological present subjunctive (ending in -Ø) versus indicative (in -a) that behaves more or less like Balkan Romance and Albanian (see §6.2.4.1.5). While aspect and mood intersect in the behavior of the perfective present and the perfective imperfect (cf. Aronson 1977), a fact that raises interesting questions for typological linguistics, from the viewpoint of differentiating mood and aspect, it is clear that the two sets of forms encode fundamental differences. The perfective is opposed to the imperfective on the basis of aspect, i.e., the quantification of the narrated event (Jakobson 1957), which in turn sometimes has modal implications when dealing with futurity, conditionality, and, insofar as it is potential, iterativity, and related concepts. Nonetheless, the subjunctive is markedly restricted to modal relations, whereas the perfective has indicative capacities, especially in the aorist, which is almost never modal (but see §6.2.4.2.8 for some modal uses). In the case of the subjunctive, modality is the primary meaning, whereas for perfectives, modality arises from the contextual restraints on quantification.
6.2.2.3.1 Macedonian
In standard Macedonian and the dialects on which it is based, the perfective present and perfective imperfect occur only in subordination to eight modal particles and related words (after Kramer 1986; see also Friedman 2002c; see Table 6.20).Footnote 241
The pfv.prs can also occur after ne ‘neg’ in questions, e.g., Zošto ne sedneš? ‘Why don’t you sit down?’
6.2.2.3.2 Bulgarian
In Bulgarian, the perfective present and the perfective imperfect can occur after the same set of particles (or their equivalents, mutatis mutandis, viz. da, šte, neka, ako, dokato [±ne], li) as in Macedonian. Bulgarian also permits these forms after what are formally or etymologically pronominal relatives (derived from WH-words) and a few conjunctions, e.g., kojto ‘who[ever], which[ever],’ kogato ‘when[ever],’ deto ‘where[ever],’ kato ‘like, as’ štom ‘as soon as.’Footnote 242 Example (6.80) illustrates an idiomatic use with kato:
(6.80)
Kato dojdeš v Ruse, šte te čakam. as you.come.pfv.prs in Ruse fut you.acc wait.1sg.ipfv.prs ‘When you come to Ruse, I will be waiting for you.’
Unlike Macedonian and Greek (but like BCMS), Bulgarian permits a free-standing perfective present, albeit rarely. Example (6.81), with bare perfectives as iteratives or potential iteratives, is illustrative:
(6.81)
Săbădja se noštem, a v glavata mi vse tova, … wake.1sg.pfv.prs refl at.night and in head.def me.dat all that.nom Zapalja cigara, podprja se na văzglavicata, i light.1sg.pfv.prs cigarette lean.1sg.pfv.prs refl on pillow.def and prodălžavam da si mislja … continue.1sg.ipfv.prs dms refl.dat think.1sg.ipfv.prs ʻIʼll wake up at night, and in my head all that stuff … I’ll light a cigarette, lean back on the pillow, and keep thinking to myself …ʼ(based on Aronson 1977: 24, citing Stankov 1969: 13)
Bulgarian contrasts with most of the other Balkan languages with regard to this intersection of superordinate aspect with mood. The differences become evident through a consideration of translations of the same text. The examples in (6.82) are translations (dates indicated in brackets) of II Timothy 4:13 ‘When you come, bring the cloak … .’ This passage is given in two Bulgarian versions as well as Macedonian, two Romani translations, Greek, two Albanian translations, and three Romanian translations, all illustrating how perfective present interacts with temporality. For Bulgarian, (6.82a) is conservative, whereas (6.82b) is more conversational. The Macedonian example (6.82c) illustrates the fact that Macedonian requires a future marker (or other modal particle if appropriate) with a present perfective. Examples (6.82d) and (6.82e) are taken from two different Romani translations of the same passage. The interesting point of these examples is that (6.82d) is translated from the Bulgarian into Bulgarian Arli Romani of Sofia, and the translator has added a future marker, as would be expected in Macedonian, but (6.82e) is from Novi Sad, in a South Vlax Romani dialect and uses a subjunctive (aves versus avesa) but kana ‘when’ suffices to condition it. The Greek in (6.82f) shows the perfective present after όταν ‘when,’ and Albanian in (6.82gh) uses the present subjunctive with the dms after kur ‘when.’Footnote 243 The Romanian (6.82ijk) shows variation between ‘when’ + present, ‘when’ + future, and a gerund.
a.
Kogato dojdeš, donesi japundžata … (Blg; Bible 1912) when come.2sg.pfv.prs bring.pfv.impv the.cloak … b.
Koga dojdeš, donesi nametaloto … (Blg; Bible 1995b) when come.2sg.pfv.prs bring.pfv.impv the.cloak … c.
Koga kje dojdeš, donesi ja nametkata (Mac; Bible 1990a) when fut come.pfv.prs bring.pfv.impv it.acc.f the.cloak d.
Kana ka aves ano mo upruni šehja (Rmi, Arli; Bible 1995a) when fut come.2sg.prs bring.impv my upper things e.
Kavadji … an kana aves (Rmi, Gurbet; Bible 1990b) cloak bring.impv when come.2sg.prs f.
Όταν έρθεις, φέρε τον φαιλόνη (Grk; Bible 1962a) when come.2sg.pfv.prs bring.2sg.pfv.impv the.acc cloakacc g.
Kur të vijësh ma bjer when dms come.2sg.sbjv me.dat.it.acc bring.2sg.impv tabaren (Alb; Bible 1980) cloak.def.acc h.
Kur të vish, bjerë gunënë (Alb; Bible 1872, 1930) when dms come.2sg.sbjv bring.2sg.impv cloak.def.acc i.
Când vii, adu şi mantaua (Rmn; Bible 1924) when come.2sg.pfv.prs bring.impv and cloak.def … j.
Când vei veni, adu-mi mantaua (Rmn; Bible 1962b) when fut.2sg come.2sg.pfv.prs bring.impv-me cloak.def … k.
Venind adu-mi felonul (Rmn; Bible 1908) coming.grd bring.impv-me cloak.def … l.
Cãndu-a s-yinji adu-nj-lu felonlu (Aro; Bible 2004) when-fut dms-come.prs.2sg bring.impv-me-it cloak.def
Bulgarian permits the future marker with ‘when’ only in an interrogative sense, e.g., Koga šte dojdeš? ‘when will you come?’.
6.2.2.3.3 Greek
Greek is midway between Macedonian and Bulgarian with regard to the boundedness of its perfective present. As in Balkan Slavic, the Greek form occurs after modal particles such as the dms να, the future θα, and the hortative ας, as well as some conditional conjunctions such as αν ‘if,’ and the modal negative imperative μη(ν).Footnote 244 Unlike Balkan Slavic, where there is a set of words which normally in Macedonian cannot but in Bulgarian must occur with the bare perfective present, some of the same or similar words in Greek occur facultatively with or without a modal particle, e.g., Grk όταν ‘when’ (see 6.83bc) and the probabilitive ίσως ‘perhaps,’ as in ίσως (να) έρθει (για) να μας δει ‘perhaps he (might) come (in.order) that he.see us’ (Householder et al. 1964: 114; Holton et al. 1997: 221–222). Like Bulgarian, Greek allows the perfective present after correlative indefinite pronouns and adverbs, e.g., όποιος ‘whoever,’ όποτε ‘whenever,’ όπου ‘where(ver),’ όπως ‘as.’ Both Aronson 1977: 22 and Holton et al. 1997: 222 note that this usage has a modal quality that is equivalent to constructions with the dms, e.g., Grk Όποιος = Ὀποιος και αν = Όποιος και να έρθει … ‘Whoever comes, [let him be greeted, etc.]’ just like the Blg Kojto = Kojto i da dojde ʻidem.’ Here Bulgarian and Greek differ from Macedonian in that Bulgarian has dedicated relativizers of the type WH+to, whereas in Macedonian, the relativizer and the interrogative are usually homonymous. Although the complementizer što can be added to specify the relativizing function in Macedonian, such forms cannot be used as indefinites; thus Blg kojto i da e = Mac koj i da e and not *kojšto i da e (see Rudin 2015 and Friedman 2015). In this respect, the Greek usage is closer to the Bulgarian. Similarly to Macedonian, however, as in (6.83a), the Greek perfective present can occur negated without the usual modal particle, as long as there is a controlling word like όταν ‘when’; as (6.83b) shows, όταν can co-occur with a modal particle, here the future marker θα. In (6.83c), temporal όταν ʻwhen’ takes the perfective present in a context where Macedonian would require the future marker and Bulgarian would use the relative form without that marker. Example (6.83d) is particularly interesting from a Balkanological point of view, since the indicative negator δεν, the modern continuation of Ancient Greek nonmodal negator οὐ (see §7.6, Footnote footnote 72), occurs in a context where Ancient Greek would have used the modal negator μή (as is still the case in modern Katharevousa Greek); it thus shows some erosion in the use of the modal negator in favor of the indicative negator, thus paralleling Balkan Slavic, where ako ‘if’ is negated by means of the same marker, ne, as occurs with simple indicative negation.
a.
όταν δεν σε δω ξανά … when neg you.acc see.1sg.pfv.prs again ‘When I will not see you again, [then I will be happy]’ b.
όταν θα σε δω when fut you.acc see.1sg.pfv.prs ‘when I will see you, [then I will be happy]’ c.
όταν σε δω when you.acc see.1sg.pfv.prs ‘when I will see you’ d.
αν δεν έρθεις if neg come.2sg.pfv.prs ‘if you don’t come’ (Katharevousa: αν μη έρθεις …)
Another difference between Greek and Balkan Slavic is in whether X or not X constructions, where Greek permits a bare perfective present, but Balkan Slavic does not (see also §4.1 and §7.7.1.3) as in examples (6.84a) (from Householder et al. 1964: 124) versus (6.84bc):
a.
έρθεις δεν έρθεις, εγώ θα φύγω (Grk) come.2sg.pfv.prs neg come.2sg.pfv.prs I.nom fut leave.1sg ‘Whether you come or not, I will leave’ b.
doagja ne doagja, jas kje otidam (Mac) comes.ipfv.prs neg comes I fut go.pfv.1sg ‘Whether he comes or not, I’m going’ c.
Peniš se ne peniš se, šte te jam (Blg) foam.ipfv.2sg intr neg foam.ipfv.2sg intr fut you eat.pfv.1sg ‘Whether you foam or not, I will eat you.’Footnote 245
The bare perfective can also occur in other Greek ‘whether X or X’ constructions, e.g., κι αν [±θα] έρθει κι αν δε [±θα] έρθει ‘whether he comes or not’ (lit., ‘and if [±will] comes and if not [±will] comes’) and the Katharevousa είτε έρθει είτε δεν έρθει ‘idem.’
6.2.2.3.4 Conclusion
For the purposes of this volume, the question is to what extent are any similarities among the relevant languages explicable as influenced by contact rather than due to typological or other factors. Most striking here is the behavior of Macedonian present and imperfect perfectives. The ability of such forms (or those of such forms that survive historically) to occur without a subordinating or modal particle is well attested throughout Slavic. Macedonian is thus unique among the Slavic languages in treating perfective presents and imperfects in a manner that can be compared to the subjunctives of Albanian and Balkan Romance. At the same time, Modern Greek and Bulgarian share certain permitted environments for perfective present where the other languages require a subjunctive. On the one hand, the explicit co-occurrence restrictions of Macedonian give the impression of being conducive to the kind of analytic explicitness favored in contact situations. On the other hand, the parallel between Bulgarian and Greek may be a reflection of contact between those two languages.
6.2.2.4 Progressive: Albanian, Turkish, Macedonian, Balkan Romance, Romani
Progressivity overlaps with, but is not identical to, imperfect and imperfective aspects. In general it can be said that the basic meaning of the imperfect is duration (as opposed to the aorist, which is nondurative), while imperfective aspect is defined as nonlimiting (as opposed to perfective, which is limiting). Progressivity focuses on the action as on-going at a given moment, and in Albanian it is compatible with both the present and the imperfect past tenses.
For Albanian, the definition in Newmark 1998: s.v. of the progressivity marker po is instructive; it “indicates momentaneous action: be {verb}ing.” The link between “action at the moment” and “on-going action” carries with it a possibility of limitation owing to its focus, and this becomes relevant in the discussion below. Albanian also has two other specifically progressive grammatical constructions: conjugated ‘be’ with the gerund (= duke/tue, etc. + participle) and conjugated ‘be’ + kah ‘to[ward]’ + present. Although Standard (at base, in part, East Rumelian) Turkish also has a so-called progressive with the suffix -I-yor; it differs significantly from the Albanian (see examples below) and the situation in WRT is also different.Footnote 246 In the case of Balkan Slavic, Balkan Romance, Greek, Judezmo, and Romani, all of these languages, independently, developed gerunds out of the inherited Indo-European present active participle in -nt-. In Greek, Macedonian, and Romani, this gerund is a genuine adverbial. In Bulgarian, it was lost and replaced by an analytic construction consisting of kato (< kak-to ‘how-relativizer) plus finite form.Footnote 247
The Balkan Romance situation is somewhat more complicated. In some Aromanian dialects, such as the Ohrid-Struga region, the gerund was also lost and replaced by analytic constructions using the deverbal noun and calqued on Macedonian (see Markovikj 2007: 122, 165). However, according to Maxim Matkartsev (p.c., 2020), the Frasheriote Aromanian dialect of Albania has calqued the Albanian progressive in po using the particle mã, and some speakers have borrowed the Albanian gerund marker duke + verbal noun. Meglenoromanian shows three types of gerunds: one borrowed from Macedonian, one calqued on Macedonian, and one inherited, albeit modified (Atanasov 2002: 235). The borrowed affix is -ḙajḱi in Tsărnarekă; in the other villages, reflexes of inherited -Vnd survive, but always followed by -Vr-a/lḙa, i.e., formants resembling the verbal noun, which is, historically, the infinitive. Finally, constructions with cu ‘with’ + verbal noun are calqued on equivalent Macedonian constructions (Macedonian so + verbal noun). In Romanian, ‘be’ + gerund still occurs as a compound tense series in the oldest documents, whereas now, indicative ‘be’ + gerund is obsolete, the meaning being expressed by various paraphrases (Zafiu 2013a: 65). Modal ‘be’ (i.e., fut, dms, cond + fi [inf]) + gerund has been reinterpreted as a kind of evidential called modul prezumtiv ‘presumptive mood’ (see §6.2.5.6). Judezmo of Bucharest shows a gerund-based progressive (e.g., sta pensandu ‘he was thinking’ (Crews 1935: 55)) but this construction continues one found in Spanish more generally.
Thus, aside from Albanian and Judezmo, no modern Balkan standard languages use the gerund to form a progressive, although dialects of Aromanian and Macedonian have developed progressive particles; see Makartsev 2020, 2021a; also Manzini & Savoia 2018: 259.
In Standard Turkish ne yapıyorsun ‘what are you doing’ (present progressive in -I-yor) corresponds exactly to Albanian ç’ka po bën ‘idem.’ In WRT, however, the progressive/gnomic opposition is collapsed, as seen in the equivalent phrase n’aparsın (using the gnomic present). In the other languages, the equivalent expression has no special marker of progressivity (see §7.4.1.2.2.1). Moreover, the Standard Turkish progressive can be used for states and habituals where Albanian cannot use a progressive (which is limited to an on-going moment in time; see §7.4.1.2.3), e.g., türkçe biliyorum ‘Turkish I-know’ versus Albanian (*po) e di shqip ‘(*prog) it I-know Albanian’ or Genelikle Debreye otobusle gidiyorum ‘Usually to-Debar by-bus I-go’ vs. Zakonisht (*po) shkoj në Dibër me autobus ‘Usually (*prog) I-go to Debar by bus.’Footnote 248
At issue, then, is the development of Albanian progressivity in its Balkan context. There are two main facts: one is the multiple realizations of the category with their distributions and the other is the surface similarity with the Slavic morpheme po, to be discussed below. As mentioned above, Albanian has three distinct realizations of progressive marking. Most Albanian dialects (both Geg and Tosk) have both po + present/imperfect and ‘be’ + gerund. A few individual peripheral dialects, both north and south, are recorded as having only one or only the other, but the distributions show no consistency except a few clusters showing only ‘be’ + gerund mentioned below. The construction present/imperfect ‘be’ + ka(h) ‘to[ward]’ + present indicative is limited to Tropoja (Nikaj-Mertur, Gash, Bytyç, Vuthaj), in northeastern Albania, and adjacent points in Malësia e Madhe (Dukagjin), northern Puka (Berisha), and southeastern Montenegro (Plava and Gusî BCMS Plav, Gusinje). In northern Dukagjin and Gash, the ka[h] construction is in competition with po but in the other regions it is the sole progressive. The po construction is the only one used in the rest of Montenegro, but this is also the case in most of Çamëri, as well as some Lab, northern Tosk (with a cluster of points in the southwest of the region), and a variety of Geg dialects (individual southern, central, northwestern, and northeastern points). The ‘be’ + gerund construction is also found uniquely in some individual Lab and Northeast Geg dialects, and there is also a cluster of such dialects for East Central Geg. Arbëresh has ‘be’ + e or çë + present indicative for this meaning, but not po, and the construction is lacking in Arvanitika (so Gjinari 2007: 391).Footnote 249
Here it is worth noting that, as reported in Makartsev 2020, 2021a, the Macedonian dialect of Boboshtica (Mac Boboščica) has copied the function of the Albanian progressive marker po into its verbal system by means of a particle gje ‘where’ + present (where gje is ultimately from Common Slavic *kŭdě) or toko (ultimately from toliko ‘so much’ – Mazon 1936: 43–44, cited in Makartsev 2020, cf. standard Macedonian tuku ‘but’ and Albanian por ‘but’ cited in the next paragraph). This is a genuine copying of function using native material – in fact related to the more common use of kaj ‘where’ with a gerundive or progressive meaning in Macedonian folklore, i.e., records of dialects prior to standardization (see §7.7.2.3.1.1). It stands in contrast to an almost literal calquing of the type conjugated ‘be’ + gerund, which can be elicited from speakers of the dialect when translating from Albanian (see Makartsev 2020). Makartsev 2021a also reports the use of ma ‘but, more, etc.’ as a progressive marker in the Frasheriote Aromanian of this same region, clearly under the influence of Albanian. Whether the usage of ma is influenced by the fact that the Albanian progressive marker po has a variety of other functions (and moreover po can mean ‘more’ in Macedonian) is not entirely clear, but there is no question that the Aromanian usage, which exists only in Albania, is somehow calqued on Albanian.
In historical terms, the progressive marker po is already attested as such in Buzuku’s text of 1555 (Fiedler 2004: 324–327) as both po and por. Meyer 1891: s.v. treats all the meanings of po and por together as coming from Latin porrō, an adverb with a variety of meanings, among which are ‘forward, further, furthermore, further on, in turn, next, moreover, on the other hand, at the same time, etc.’ In modern Albanian, por means ‘but,’ although the Turkism ama is more common colloquially. Po has three basic types of meaning: (1) confirmative ‘yes, indeed, precisely, exactly, etc.,’ (2) conditional ‘if’ (when followed by the dms të or the aorist), and (3) progressive.Footnote 250 Skok 1972, s.v. cites po in the BCMS dialect of Kosovo as meaning ‘but, perhaps, and so’ as a borrowing from Geg. Balto-Slavic po, which can be a preposition, preverb, comparative marker, etc., is much older and has a variety of cognates in various Indo-European languages. Among the many meanings of po as a preverb in Slavic is ‘to do for a little while,’ and this has a potential connection with the focus of Albanian po on the moment of an action in progress. It is telling, however, that in some of the regions with the most intensive Slavic contact, progressive use of po is lacking and only the kah or gerund constructions occur. Given the distribution of po(r) among Geg and Tosk dialects combined with the situation in Arbëresh and Arvanitika, it would seem that the specification of progressivity is peculiar to Albanian in its Balkan context. Especially striking in this respect is the obsolescence of such constructions in Romanian and of the gerund in Bulgarian and SDBR.Footnote 251
It should be noted, however, that progressivity gives Albanian a nuance that coincides with some Macedonian derived imperfectives in -uva. As Humphries 1997: 91 shows, with some verbs, derived imperfectives in -uva focus on the moment of speech, e.g., taa sekogaš mi nalaga… ‘she always orders me [to do something or other]’ (a habitual problem) versus taa mi naložuva… ‘she is ordering me [to do this] (right now!).’ Moreover, the Albanian ethnolect of Macedonian in North Macedonia shows a tendency to expand -uva where Macedonian first-language speakers would not use it, e.g., dojduva (standard doagja) as the equivalent of po shkon ‘s/he is going.’ Such usage, however, is spreading among the younger generation of Macedonian first-language speakers, where, for example, the aspectual pair izloži/izlaga ‘exhibit.pfv/ipfv’ has been replaced by izloži/izložuva (which latter would be marked as habitual in the older norm) in the speech of the current youngest adult generation (Markovikj 2017). Although it is generally recognized that imperfective aspect is unmarked vis-à-vis perfective, it is arguable that the highly productive use of -uva to derive imperfectives in Macedonian that contrast to underived imperfectives as well as perfectives suggests a convergence that also involves Albanian progressive marking as a kind of marking of imperfectivity. Here the fact can be stressed that Macedonian -uva is more productive than in any other Slavic language, and, moreover the “overuse” of -uva by native speakers of Macedonian is a site of anxiety for prescriptivists.
To the extent that some Romani dialects with Turkish conjugation (see Tables 6.17–6.19) borrow both the gnomic present in -r and the progressive present in -y(or), this opposition has entered those dialects. It is worth noting here that some Romani dialects borrow only one or the other Turkish present and use it in all contexts where Turkish itself would make an aspectual distinction (Friedman 2013b; StTrk -yor is sometimes realized as WRT -y).
6.2.3 Taxis and Resultativity: Pluperfect and Perfect
In this section we accept Jakobson’s 1957 definition of taxis as specifying the relationship of two narrated events. Resultativity denotes the category encoding the result of an event, i.e., a resultative refers to a state of affairs that results from some previous event. The difference between taxis and resultativity can be seen in standard Macedonian and the western central dialects on which it is based, where the inherited pluperfect using the imperfect of ‘be’ plus the verbal l-form (the old resultative participle) and the new pluperfect formed with the imperfect of ‘have’ plus the neuter verbal adjective (descended from the old past passive participle) are in competition. The former is marked for pastness and taxis while the latter is marked for pastness and resultativity. Thus for example the difference between Toj mi ja pokaža, no jas vekje ja bev videl and Toj mi ja pokaža, no jas vekje ja imav videno – which could both be translated ‘He pointed her out to me, but I had already seen her’ – is that bev videl means I had just spotted her before he did (e.g., crossing the street), whereas imav videno means I had seen her on some previous occasion (and, e.g., already knew she was in town). The former thus relates an event temporally to some preceding event without implying a resultant state, whereas the latter specifies a resultant state from the preceding event (cf. Friedman 2014b: 94).
As Benveniste 1952 argued, there is a well-established semantic connection between possession and perfect. This accounts for such constructions as in North Russian dialects (u menja hoženo ʻI have gone’ – lit., ‘by me go.pass.ptcp.n’) as well as the oblique subject in the Armenian perfect, e.g., nora bereal ē ʻhe has carried’ (lit., ʻof.him carried is’). Nonetheless, the developments with ‘have’ perfects in the Balkans have clearly taken place in the context of convergent contact. Moreover the use of ‘be’ as an auxiliary for the perfect also has some areal implications in the Balkans. This section examines the issues of perfect formation in the Balkans by examining relevant pairs of languages and language groups. Very briefly put, the impetus for ‘have’ perfects in the Balkans appears to be Romance (for Albanian, Greek, and Balkan Slavic), and, secondarily, at the local level, Greek (for Balkan Slavic, Romani), while the spread of ‘be’ as an auxiliary in Balkan Romance can be attributed to Balkan Slavic. At the same time, the effect of Albanian at the local level (for Macedonian) cannot be ruled out and even has some interesting support (e.g., lexical items as well as some productive morphemes and stress patterns in some Kostur dialects; see Friedman 2018a). See also Makarova 2021 on Albanian, Aromanian, and Macedonian in the Prespa basin, and Adamou 2012c on multiple causations for the developments. For the most part, Judezmo and West Rumelian Turkish do not enter into consideration here, except for auxiliary usage in the latter (and also some Arli Romani). This section is organized by individual language groups and discusses the outcomes of contact in each group.Footnote 252
6.2.3.1 Albanian
The Albanian ‘have’ perfects (‘have’ + participle) are already in place by the time of our earliest extensive attestation (1555). The auxiliary ‘be’ is used for mediopassives. In Northern Geg, however, ‘be’ is used to form the perfect of ‘be’ and in Northeast Geg also for verbs of motion. The similarity to Western Romance has led to the hypothesis of Romance influence here, as with the Geg ‘have’ future, and indeed in Kosovo the consistent use of both as well as the earlier attestations of Romance speakers into the twentieth century is worth noting.Footnote 253 Given the massive lexical impact of Latin on Common Albanian (i.e., the stage before the Geg/Tosk split), and given that the ‘have’ perfect in Latin dates from the second to first centuries BCE, i.e., precisely the period when Rome was annexing the Balkans, it is not unreasonable to hypothesize that the Albanian analytic perfects arose during the period of contact with Latin. It is worth observing here that by the time Latin arrived in the Balkans, the synthetic perfect in Greek was already in retreat before the aorist on the one hand and, less so, an analytic ‘have’ or ‘be’ perfect with a participle on the other (Chantraine 1926; Horrocks 2010: 131). We can also note here that the flexibility of auxiliary placement (usually proclitic but on occasion enclitic, especially in the North, see Boretzky 1966) points to a relatively recent formation, i.e., one that does not stretch as far back in time as the period before contact with Latin. Unfortunately, unless older texts come to light a definitive answer is impossible.
Albanian also has compound perfects and pluperfects, which use the (analytic) perfect and pluperfect of ‘have’ or ‘be’ as auxiliaries. See Table 6.30 and Footnote footnote 313 for illustrative formations and examples (6.85–6.88) for usages. Note that the short form participle pasë rather than the long form pasur is used for ‘have.’ The spelling pas is colloquial (cf. English nite for night). These constructions are typical of Geg (northern), but they are also found in Lab (southern – Totoni 1971: 73; Gjinari 1989: 250) and can be used to render an extra degree of anteriority. Example (6.85) is illustrative: a past event is indicated with an aorist, an event before that with a pluperfect, and an event prior to the second one with a compound pluperfect (key elements in bold):
(6.85)
Është e vetmja brengë, që më mbetet – shqiptoi më qartë ai, pasi kishte folur një copë herë, në mënyrë të ngatërruar, për një vajzë të bukur dhe inteligjente, me të cilën e kishin pasë fejuar prindët qysh në fëmijëri (Sh. Demiraj 1976: 271) ‘“It’s the only trouble I have.” – he said [aorist] more clearly, after he had spoken [pluperfect] in confusing manner for some time about a beautiful and intelligent girl to whom his parents had engaged [compound pluperfect] him in childhood.’
It can also be used as a distant past habitual, e.g.,:
(6.86)
kam pas lexuar shumë libra të tillë have.1sg had.ptcp read.ptcp many books such ‘I used to read a lot of books like that’ [implied: ‘a long time ago’ or ‘but not anymore’].
Such tenses can also be used for jocular effect, as in the following extended example (key elements in bold):
(6.87)
Lul: “Gjysh, tregoma një fjali në kohen e shkuar e të tejshkuar.” Tafë: “Shkruaje … Shkruaje … Na kemi pas pasur kafe” (Rilindja 82.II.3: 8). ‘L. [holding a homework assignment] – Grandpa, tell me a sentence in the past pluperfect T. [thinking] – Write … Write … [stating] We used to have coffee.’
Adding to the effect is the fact that the tense is misidentified in terms of standard Albanian. In the standard, the term e kryera e tejshkuar ‘distant past perfect’ refers to an ordinary analytic pluperfect using the aorist rather than the imperfect of the auxiliary (e.g., pati pasur rather than kishte pasur for ‘had had’). The compound pasts are labeled ‘secondary’ or ‘compound’: e kryera e dytë ‘second perfect,’ më se e kryera (or e kryera e plotë) e dytë ‘second pluperfect’ or forma të mbipërbëra të së kryerës/më se të kryerës ‘overcompound form of the perfect/pluperfect’ or kohët e përbëra ‘the compound tenses.’ In much of Geg, however, analytic perfects replace aorists, especially as auxiliaries. In Luzni, on the eastern edge of West Central Geg, Beci 1974: 250 has even recorded a mediopassive double compound perfect, i.e., the perfect of ‘have’ used as an auxiliary with the participle of ‘be’ plus the main verb (also a participle):
(6.88)
Kan pas qõn bõ have.3pl had.ptcp been.ptcp made.ptcp ‘they had become’ or ‘they have been made’ (lit., ‘have had been done’).
The existence of the compound pasts in the Lab dialect has also allowed the short participle into the standard language, albeit in a very marginal role.
6.2.3.2 Balkan Romance and Balkan Slavic
When Slavic arrived in the Balkans, the perfect was formed by the present imperfective of ‘be’ plus the resultative participle. This perfect developed into the unmarked past throughout Slavic.Footnote 254 Possessive perfects using ‘have’ (or, in the case of Russian, possessive constructions) have developed at three Slavic peripheries: in Czech, Polish, and Kashubian in contact with German, in North Russian dialects in contact with Finnic, and in Balkan Slavic in contact with Balkan Romance, Albanian, and Greek. The Balkan Slavic constructions do not appear in the written record until the early modern period, but this may be a result of the limitations on the sources of earlier records. In any case, the Macedonian ‘have’ perfect appears to be contact-induced. In most of Bulgarian, such constructions are limited to animate subjects and transitive verbs (true past passive participles), which agree with a direct object, except in Thrace, where, as in standard Macedonian and the western dialects on which it is based, the old past passive participle has become a true verbal adjective (formed from intransitives as well). In the ‘have’ perfect in those dialects, the verbal adjective usually agrees with the direct object for transitives and the subject for intransitives, although some invariant neuter verbal adjectives have been recorded for the Malgara region (Bojadžiev 1991: 93).Footnote 255 For Balkan Romance and Balkan Slavic, as Gołąb 1976, 1984a: 135 has convincingly argued, there is a split between East Balkan Slavic influence on North Danubian Romance and SDBR influence on West Balkan Slavic (extending along the Via Egnatia into East Balkan Slavic). At issue are the use of ‘be’ and ‘have’ as auxiliaries, the nature of the participle they accompany, questions of paradigmatic versus syntactic status, additional extensions of analytic paradigmatic constructions, and variability in directionality.
In terms of the pluperfect, it is worth noting that while Bulgarian has expanded the use of ‘be’ to include marked evidentials (see §6.2.5.1), and the inherited pluperfect using the imperfect of ‘be’ and the resultative participle is quite frequent for the indication of taxis, in Macedonian, the inherited pluperfect (expanded to include both aorist and imperfect l-form stems) is well on the way to obsolescence. Although the oldest generation still has a sense of how to use the form, it is unfamiliar to the youngest generation, who either use a taxic form of the ‘have’ perfect or depend on context to situate an aorist functioning as anterior to another aorist.Footnote 256
6.2.3.2.1 Macedonian and Aromanian
While each of the main non-Slavic contact languages (Albanian, Greek, Balkan Romance) have been proposed as the source for the BSl ‘have’ perfect, Gołąb’s 1976, 1984a arguments combined with the geographic distribution make Aromanian the most likely source for Macedonian, whereas in Thrace, the causation is harder to tease out: on the one hand, Greek was a colloquial language used in the countryside (which was not the case in Macedonia, except in the southernmost districts of what is today Greek Macedonia), while on the other hand, Macedonian and Albanian speakers migrated to Thrace during the early modern period, from precisely the regions where there were also many Aromanians. We can also note here that the Via Egnatia ran through this part of Thrace and could have served as a source of Latin influence, given the posting of Roman soldiers along the route. For Macedonian, Gołąb points out that the Aromanian ‘have’ perfect uses an invariant feminine verbal adjective, and since feminine functions as the unmarked (neuter) gender in Aromanian (as in Albanian), it corresponds to the invariant neuter verbal adjective in Macedonian. Moreover, the ‘have’ perfect of ‘be’ and ‘have’ (1sg imam bideno, imam imano) occur only in the Ohrid-Struga region in southwestern Macedonia and further south and southwest in the Kostur and Korça regions. As one moves north and east from this center of innovation, ‘have’ perfects become less frequent, with fewer paradigms (i.e., the presence versus absence of the use of the imperfect and old perfect of ‘have’ as auxiliaries), and before World War Two, they did not occur north or east of the Vardar.Footnote 257 Additional evidence for precisely Macedonian–Aromanian interaction is the fact that in southwestern Macedonia, the two perfect systems have become isomorphic, as illustrated in Table 6.21 (based on Gołąb 1984a: 135).
| Macedonian | Aromanian | ||||
|---|---|---|---|---|---|
| I | sum večeral | amu tsinátə | |||
| II | sum večeral | imam večerano [borrowed] | < | amu tsinátə | |
| III | imam večerano | sum večeral | > | ésku tsinátu [borrowed] | amu tsinátə |
| IV | imam večerano | sum večeran [reborrowed] | < | ésku tsinátu | amu tsinátə |
Stage I represents the situation when Slavic and Romance speakers first came into contact in southwestern Macedonia. Slavic had a perfect using ‘be’ and a resultative participle that agreed in gender/number with the subject. Romance used ‘have’ plus an invariant feminine past passive participle, since feminine was the unmarked gender, the neuter having been lost. In Stage II, Macedonian calqued the Aromanian construction by using ‘have’ and a neuter past passive participle (which became a verbal adjective, formable from intransitive verbs). In Stage III (which could have overlapped with Stage II), Aromanian calqued the Macedonian perfect by using ‘be’ and a past participle that agreed with the subject. (According to Makarova 2021, this calque occurs only in what is now North Macedonia but not in what is now Greece and Albania.) In Stage IV, Macedonian re-calqued the Aromanian calque by using ‘be’ plus the verbal adjective and agreeing in subject.Footnote 258 Most of western Macedonian retained the old perfect (‘be’ + l-form) of Slavic, which took on nonconfirmative meanings – in contrast to the markedly confirmative synthetic aorist and imperfect – while still retaining some of its perfect usages. As one moves further southwest, the old l-perfect becomes strictly limited to evidential contexts, while the new ‘be’ perfect can only describe a present result. Thus, an adverbial like tri saati ‘for three hours’ can be used with imam večerano but not with sum večeran, both of which can be translated by English ‘I have dined.’ The same restrictions apply in Aromanian. Moreover, in the regions where ‘be’ and ‘have’ can be used in the ‘have’ perfect (1sg imam bideno, imam imano), the l-perfect is never used except in marked evidential (nonconfirmative) contexts. The situation is thus most congruent with the neighboring Albanian dialects (and the standard language, which is based on them), where the admirative is a marked nonconfirmative (see Friedman 2012b). On the other hand, the complete absence of the synthetic pluperfect in SDBR and its replacement by the analytic one can be seen as the result of contact with Macedonian, Albanian, and Greek (depending on the region).
6.2.3.2.2 Bulgarian and Romanian
As indicated above, in the competition between ‘be’ and ‘have’ as auxiliaries involved with resultativity and related concepts that depend on one event being prior to some other event or state, e.g., taxis, it is clear from the historical record that Latin brought ‘have’ to the Balkans, whereas Slavic brought ‘be’ in these same functions. Gołąb 1976 argues that whereas the effect moves from Aromanian to Macedonian as described in §6.2.3.2.1, in the case of North Danubian Balkan Romance, the direction of influence is from Bulgarian to Romanian. This is seen not only in the elaboration of ‘be’ as the auxiliary for new paradigms in Bulgarian (thus, for example, where Macedonian has imal napraveno Bulgarian has bil napravil for ‘he had/has supposedly done’) but also in the use of ‘be’ in Romanian for active compound tenses, e.g., the perfect of ‘be’ forming a pluperfect, e.g., am fost făcut ‘I had done’ and invariant fi in compound tenses such as the anterior future, e.g., o fi făcut ‘he will have done’ corresponding to Bulgarian bil săm napravil and šte săm napravil, respectively.Footnote 259 As Gołąb 1976: 306 observes, the use of a fi with a (transitive) perfect participle (făcut) should be passive; in Bulgarian however, the participle napravil is etymologically resultative, and thus ‘be’ carries a different force. What we have, then, is a surface reinterpretation of ‘be’ as a past-forming auxiliary and an equivalence of the two forms of participial origin used to form past tenses (in Bulgarian, the form retains participial features, but not in BCMS or Macedonian). Gołąb further argues that the Old Slavonic perfect was not calqued into Romanian because habeō factum (> am făcut) was already well established in Common Romance. The other analytic tenses, however, were not well established, whence the calquing.Footnote 260
Note that the Aromanian situation shows more mutuality, in keeping with the relative social positions of local populations. Gołąb 1976: 304 hypothesizes on the basis of the available evidence that north of the Danube the landowners (boieri) and transhumant shepherds (and eventually shepherds who moved down from the mountains and became peasants) spoke Romance and the (initially settled) peasantry spoke Slavic; Slavonic, however, was the language of literacy. South of the Danube, Slavic was the language of the landowners (bojari) and clergy as well as the peasantry, and the Romance-speaking population accommodated by shifting, unless they stayed isolated up in the mountains. Thus we have a situation where Romanian shows significant influence from Bulgarian, and Macedonian from Aromanian, while Bulgarian and Aromanian themselves remain less affected. Aromanian, however, was more affected by Macedonian than Bulgarian was by Romanian, itself a possible artifact of the linguistic situation at the time of Slavic settlement.Footnote 261
6.2.3.3 Greek
The Modern Greek ‘have’ perfect followed a path of development quite distinct from that of the languages discussed so far. As noted in §6.2.3, the use of the synthetic perfect was in decline in the Hellenistic period and the form and category had already begun to merge with the aorist (Chantraine 1926; Horrocks 2010: 102), although contact with Latin may have contributed to the already existing tendency (Horrocks 2010: 131). Medieval and Modern Greek have two ‘have’-based perfect formations, one consisting of ‘have’ as an auxiliary with a past passive participle, e.g., έχω γραμμένο(ν) ‘have.1sg write.pst.pass.ptcp’ and one consisting of ‘have’ plus the remnant of the old aorist infinitive, e.g., γράψει (from earlier γράψειν, with an ending from elsewhere in the infinitival system replacing Classical Greek -αι), now a nonfinite form confined to use in the perfect system. The former type is earlier whereas the latter type is much more recent. It is a development out of a ‘have’-based analytic future of Hellenistic and early Byzantine times, consisting of ‘have’ plus an infinitive, undoubtedly calqued on a Late Latin model, which with a past tense form of ‘have’ was used as a future-in-the-past. From that meaning, it came to serve in the apodosis in a conditional sentence (cf., e.g., King James’s English If you had been here, my brother had not died [John 11:21]). This combination then became interpreted as a true (i.e., indicative) pluperfect.Footnote 262 The earliest example of such a ‘have’ pluperfect (i.e., past perfect) is from the fourteenth century (Chronicle of Morea), whereas the specifically present perfect form of the ‘have’ perfect is a later, modern development, analogically based on the pluperfect. It is not mentioned, for instance, in various sixteenth and seventeeth century grammars of contemporary Greek; thus Nikolaos Sophianos’s grammar from c. 1555 (most recent edition: Papadopoulos 1977) has as the only present perfect form, the ‘have’ plus passive participle type (γραμμένον ἔχω), but for the pluperfect gives both that type and the infinitival type (thus both γραμμένον εἶχα and γράψει εἶχα, while the grammars of Girolamo Germano (1622) and Simon Portius (1638) both show (e.g., Portius, p. 70) alongside a ‘have’-based pluperfect, e.g., εἶχα γράψει ‘I had written,’ as the pluperfect (with εἶχα γραμμένα as a variant), the aorist ἔγραψα ‘I wrote / I have written’ as the equivalent of the present perfect. Moreover, the ἔχω γράψει type was still rare in the nineteenth century (Horrocks 2010: 297–301; Joseph 2000a; Thumb 1912: 162–163). Joseph 2000a, fleshing out the account in Thumb 1912: 162, attributes the shift from modal to indicative pluperfect to internal developments within Greek whereas Horrocks 2010: 346–347 sees post-1204 contact with Romance as responsible.
Turning now to the perfect construction using various participles, this is already attested in Classical times with ‘be’ as an auxiliary, and it spread in Hellenistic times; contact with Latin and, after 1204, Romance could have strengthened these already existing tendencies (Horrocks 2010: 131–132). The constructions with ‘have’ + participle, especially perfect passive participle in a transitive sense, are probably due directly to contact with Latin (Horrocks 2010: 131). In the ‘have’ + participle construction, the participle generally agrees with the direct object, e.g., την έχω ιδωμένη ‘I have seen her’ (lit., ‘her.f.acc have.prs.1sg see.ptcp.f.acc’). However, in various dialects, including ones in contact with Macedonian, Albanian, and Aromanian, the participle can be invariant -o, which is both masculine (acc) and neuter singular (nom-acc) or -a, which is neuter plural (nom-acc). Although ‘have’ + participle is still a possible expression of resultativity in Modern Greek, it is not used with intransitives; e.g., with φταρνίζομαι ‘sneeze’ only the ‘have’ + the former aorist infinitive construction is acceptable. The ‘have’ + participle construction was in competition with the ‘have’ + former aorist infinitive construction, and dialectally one or the other was preferred (Thumb 1912: 161–162), but the evidence presented above indicates that the two constructions still differ.
The use of invariant neuter plural in the past passive participle in Greek, attested in Simon Portius’s 1638 grammar, suggests an influence of Albanian, which can also be seen in the admirative of the Frasheriote Bela di Suprã (Mac Gorna Belica) dialect of Aromanian (see §6.2.5.6 below, and Friedman 1994b). It is here that the observations of Sandfeld 1930: 105–106, 115 are relevant and must be presented but also critiqued. Sandfeld 1930: 106 assumed that Greek had to be the source of the perfect in Albanian, Macedonian, and Balkan Romance. Both the chronology and the details, however, do not substantiate this, and in fact they support the view that Greek was influenced by contact with the various other languages. Of particular importance here, in addition to the data adduced above, is the desinence of the relevant participle in each of the languages. The Albanian participle ended in unstressed schwa (orthographic -ë) from older participial *-(n)o-. The schwa was preserved in (southern) Tosk, but not in the standard language (except in monosyllabic stems, e.g., dialectal punuarë/larë, standard punuar/larë ‘working ~ worked/washing ~ washed’). In Albanian adjectives, when -ë is added to adjectival stems ending in a consonant, it signals masculine plural. Meanwhile, in Aromanian, as noted above, the final schwa in the participle in the perfect is feminine singular. It is here that the Frasheriote Bela di Suprã admirative becomes relevant. In those forms, it is a masculine plural participle (usually imperfect) that is the base (see §6.2.5.6). What we see, then, is the association of masculine plural with participially based paradigms in Aromanian calques on Albanian. In the instance of the Greek dialects with invariant neuter plural participle, which in and of itself is inconsistent with the general trends of Greek morphosyntax in not agreeing with anything, the fact that the ending is -α, the closest Greek sound to schwa, suggests that the influence was into and not out of Greek. This is especially relevant when we keep in mind that most of Attica was Arvanitika-speaking for centuries, and that the Arvanitika dialects preserve(d) the pronunciation of final schwa to this day. In fact, Thumb’s one example (§227, from Texts III.4), την έχω ιδωμένα ‘I have seen her᾽(lit., ‘her.acc.f have.prs.1sg see.ptcp.n.pl,’ i.e., ‘I have seen her,’ rather than the more usual την έχω ιδωμένη (with feminine, as above), is from Aegina, an island near Athens where Arvanitika used to predominate. When examined in the light of what we know about the histories of Albanian, Aromanian, and Macedonian, then, the Modern Greek paradigmatic perfect emerges as another example of Romance influence, while the participial construction, in addition to the obvious Romance impetus, suggests further influences from Albanian.
Particularly noteworthy for Greek is the observation of Householder et al. 1964: 134 that the perfect and pluperfect are rare in both spoken and written Greek, at least for the mid-twentieth century period of Modern Greek they were describing (and note the absence, mentioned earlier in this section, of a present perfect from the early 17th century grammars of Girolamo Germano and Simon Portius). For the indicative, the aorist is normally used when the context is clear, while in the conditional θα + imperfect is used. In their corpus of 2,138 verb forms, there were 381 plain imperfects (eight with θα, fourteen with να), 451 plain aorists (none with θα, two with να), eight plain perfects (none with θα, one with να), thirty-five plain pluperfects (one with θα, none with να);Footnote 263 cf. §6.2.3.2 and §6.2.4.2.2. Moreover, hortative ας occurred only twice, both times with a present. These figures suggest that unlike the situation in Balkan Romance, Balkan Slavic, and Albanian, the Greek perfect is extremely marginal. This in turn emphasizes the connection among the first three languages and also the fact that the fourth is a separate and later development (and secondary to the pluperfect, again, unlike the first three).
6.2.3.4 Greek and Romani
Historically, Romani, like its Middle Indic ancestor, was a ‘be’ language, as is clear from the evidence in Paspati 1870: 98; cf. also Matras 2002: 174. In general, the simple preterite also conveys perfect meanings in context. Some dialects in contact with Greek use the root ther- (etymologically ‘hold’) as ‘have’ and form a perfect with it using the Romani participle, e.g., ov therel našto ‘he has left’ (Matras 2004: 88). This usage is notably absent in Romani dialects elsewhere in the Balkans.
6.2.3.5 Judezmo
Bitola Judezmo uses both inherited tener ‘hold’, now meaning ‘have,’ and inherited aver ‘have’ plus the past participle to form the perfect, e.g., tengu favladu and a favladu ‘I have spoken’ (see, e.g., Malinowski 1984). The latter is used less frequently (Luria 1930: 132). Tenere was used in Asturian and preferred in Portuguese and Mirandese. Gabinskij 1992: 85, 97–98 writes that tengo is preferred in Judezmo in general, e.g., tengo estado, tengo ido ‘I have been, I have gone,’ tyene komido tres mansanas ‘he has eaten three apples’ (with no agreement). The Judezmo analytic pluperfect is more common than or completely replaces the synthetic pluperfect. Some Judezmo dialects have completely lost the older analytic perfect in ‘have’ (aver) and replaced it with a newer one in ‘hold’ (tener). Such usages, however, are also attested in Asturian, Mirandese, and Portuguese, with the former having more or less eliminated the simple preterite. There is a tendency in Judezmo to blur the distinctions between the two types of pasts (Gabinskij 1992: 96). See §§6.2.1.3. 6.2.3 for additional discussion of perfects. From a Balkanological point of view, these developments could be explained as the continuation of inherited tendencies, but the maintenance of a ‘have’ perfect is also consistent with the contact environment.
6.2.3.6 Turkish
Like Romani, Turkish is a ‘be’ language with no lexical ‘have.’ Possessive-based futures accordingly calque ‘have’ constructions (cf. §6.2.4.1.5 on Romani, §6.2.4.1.6 on Turkish). The only Balkan feature in this regard is the WRT tendency (shared with the Armenian-influenced Turkish of eastern Anatolia [Lewis 1967: 119]) to separate enclitic forms of ‘be’ rather than agglutinating them, e.g., gelmiş idim versus gelmiştim ‘I had come.’ We can also mention here those Romani dialects with the heaviest borrowing of Turkish conjugation (Tables 6.17–6.19): those that have borrowed copular idi ‘3sg.pst’ have the potential of forming Turkish pluperfects for verbs using Turkish conjugation. Note, however, that for the most part, Romani uses the native remoteness marker -as or the preterite (s)ine ‘3sg.pst’ with the perfective past to form pluperfects.
6.2.4 Mood: Futures, Conditionals, and Volitionals
The view of modality employed here is that of Kuryłowicz 1956: 26, Gołąb 1964b, Lyons 1969: 304ff., and Aronson 1977, among others, who define mood in terms of the ontological qualification of the narrated event as real or unreal. According to this analysis, futures are considered with conditionals and volitionals (cf. Ammann & van der Auwera 2004), a treatment that is also consistent with their respective historical developments.Footnote 264 There are two main issues relevant here. One is the competition between ‘want,’ ‘have,’ and ‘be’ as auxiliaries; this is especially important in the development of the future, but also has relevance for the conditional. The second is the reduction of finite forms to particles (or affixes), a trait present to some extent in all the Balkan languages, but to varying degrees. We can note that the freedom of word order in the auxiliary is related to its degree of reduction to a particle and the degree of infinitive loss. As pointed out in Matras 2007, modal forms are especially amenable to contact-induced change. The most important of the Balkan modal markers is the dms (Friedman 1985a, and see §4.3.3.1.2, Footnote footnote 145), which is discussed at length in §7.7.2.1.3.1 in connection with the replacement of the infinitive.
6.2.4.1 Futures: An Overview
When Slavic entered the Balkans (sixth to seventh centuries CE), the synthetic futures of Ancient Greek and Classical Latin were already obsolete and Slavic itself (based on the evidence of OCS) had no systemic future. There was competition between the auxiliaries ‘have’ and ‘want’ + infinitive to mark futurity in Latin and Greek, with Latin ultimately favoring ‘have’ (seen all across the Romance languages) and Greek favoring ‘want.’ OCS used the perfective of ‘be’ in addition to ‘want,’ ‘have,’ and various forms of ‘begin’ + infinitive. The ‘want’ + infinitive construction survives (with modified or new infinitives) in Romanian, in Northwest Geg (near and in Montenegro), in Bulgarian (with postposed auxiliary) as an archaism or dialectism, and Meglenoromanian (for speculations and threats). This form also survives in all the non-Balkan Štokavian and Čakavian dialects of BCMS and connects them with East South Slavic. In fact, much of Štokavian ended up in its current location as a result of northward migrations during the fifteenth to eighteenth centuries. The rest of Slavic, including Kajkavian, which goes with Slovene in this respect as in many others, developed the perfective of ‘be’ as a future marker.Footnote 265 The next stage was ‘want’ + dms + conjugated present tense verb (for Greek in the fourteenth century, for Slavic the fifteenth century). This stage also survives in BCMS, including Torlak dialects. The third stage, which overlaps the second, is the transformation of ‘want’ into an invariant particle + dms + conjugated main verb. This type of construction is still the main one in Tosk (and Standard Albanian) and in parts of Geg. It also occurs in northern Aromanian (Papahagi 1974: 70); it is characteristic of southern Romanian and survives in Torlak BCMS and in certain modal uses in East South Slavic and Romani, but not in Greek. The fourth stage is the elimination of the dms so that the future is marked by an invariant particle plus a conjugated verb. In addition to being the standard future in Balkan and southern Vlax Romani, Greek, and Balkan Slavic (including Torlak BCMS), it is very common in colloquial Tosk (including informal written Standard Albanian) and southern Aromanian (Papahagi 1974: 70). In Meglenoromanian, the original future marker merged with dms, producing a new particle, ãs, in Tsărnarekă, but eliminating a distinct future marker in the other villages, where the dms alone (si or sã) doubles as a future marker (Atanasov 2002: 249).Footnote 266 Romani outside the Balkans has other means of forming or expressing the future, and it appears that the Romani development in the Balkans occurred in concert with the other Balkan languages.
The conjugated ‘have’ future (+ dms + subjunctive) is one among several standard variants in Romanian and can be encountered in Meglenoromanian (Atanasov 2002: 248). It also occurs in the Arbëresh of Calabria and Sicily (but not further north and east, where ‘want’ is used; Gjinari 2007: Map 305). Conjugated ‘have’ + infinitive (most of Geg) or + për + deverbal nominalization (a few Tosk dialects) occurs in Albanian. Invariant negative ‘have’ (= negative existential) + dms + conjugated.verb is normal for Balkan Slavic and also for dialects in contact with it (Romani, WRT, Aromanian), while positive ‘have’ (both invariant and conjugated) has various modal connotations (Bužarovska & Mitkovska 2019).
Thus, while the historical record makes it clear that the seeds of the Balkan future were already present in Latin, Greek, and Slavic at the times of contact, and while the lexical sources of the auxiliaries are typologically ordinary, and while the specific sets of structural changes can each be explained language-internally, nonetheless it is equally clear that the ‘want’ future in the Balkans is an example of mutual reinforcement and feature selection under conditions of language contact that began to take shape in the late middle ages but did not reach its current state until the early modern period, and in some areas, e.g., parts of Albania, and in Romani dialects, the process is still on-going.Footnote 267 Moreover, competition with the ‘have’ future also shows local variation. Thus, while similar types of futures have developed elsewhere, the evidence of Balkan Slavic, Balkan Romance, and Balkan Indic (Romani) make clear the fact that in the context of European history and geography the Balkan future is indeed a Balkanism. Moreover, regardless of where the centers of diffusion might originally have been located, in the more recent past the intersecting linguistic peripheries of western Macedonia and adjacent parts of Albania emerge as a center of innovation. In looking at the expression of future constructions, we see, on the one hand, that ‘want’ is still spreading at the expense of ‘have,’ ‘have’ is not altogether vanquished, and reduction to an invariant marker is not altogether complete, especially in Romanian, BCMS, and Bulgarian (especially for the Balkan conditional, where the auxiliary conjugates).
6.2.4.1.1 Greek
For Greek, ἔχω ‘have’ and μέλλω ‘be about to’ + infinitive emerged as the main competitors with the synthetic future during the early Postclassical period, with θέλω ‘want’ – even though attested (sparsely) in Postclassical Greek as early as the sixth to seventh centuries in papyri and with earlier antecedents in Classical and Hellenistic Greek (Holton et al. 2019: 1781) – not supplanting μέλλω and ἔχω until the Middle Ages.Footnote 268 As Holton et al. 2019: 1781 put it, “Future-referring θέλω + infinitive can already occasionally be found in A[ncient]G[reek] and the lower registers of Hellenistic Greek … but its breakthrough is probably to be dated to the E[arly]Med[ieval]G[reek] period,” a stretch of time covering c. 500–1100, and thus a chronology that is Balkanologically significant, given the presence of Slavs in the Balkans from about the sixth century, and Albanians and Romans from considerably earlier.
The ἔχω type was most probably influenced by, if not completely based on, the Late Latin infinitive + habēre ‘have’ future that may have played a role as well in the Geg Albanian future and the Balkan Romance future, and thus ultimately the Greek perfect via a conditional sense, as discussed in §6.2.3.3. Horrocks 2010: 300 attributes the victory of ‘want’ over ‘have’ as the future marker to this development of ‘have’ as the perfect marker. It has been claimed that the θέλω + infinitive future was consistently distinguished from θέλω + νά + finite verb to express volition, but that may have held only for “the first centuries of the L[ate]Med[ieval]G[reek] period” (c. 1100–1500), as noted by Holton et al. 2019: 1788; Joseph & Pappas 2002 document clear instances of future θέλω + νά + finite verb constructions in Medieval Greek, and Holton et al. 2019: 1788 also offer several examples. The θέλω + infinitive future continued to be used – at least in texts – until the infinitive disappeared in the sixteenth century, although Thumb 1912: §226 reports it dialectally into the nineteenth century.Footnote 269
Meanwhile, two different paths of development emerged from the θέλω + infinitive future, both with Balkanological import. The path that led to the widespread modern future marker θα took the following form, using the verb γράφω ‘write’ as an example of a main verb, shown here in its imperfective form; we gloss θέλω as ‘want’ in the topmost line as that was the starting point but as ‘will’ thereafter to indicate the futurity of the combination (Table 6.22).
| θέλω γράφειν | ‘want.1sg write.inf’ |
| ⇓ | |
| θέλω να γράφω | ‘will.1sg dms write.1sg’ (by on-going replacement of infinitive (see §7.7.2.1.1.2.1)) |
| ⇓ | |
| θέλει να γράφω | ‘will.3sg dms write.1sg’ (by elimination of redundant person/number marking on auxiliary verb, giving an impersonal auxiliary)Footnote 270 |
| ⇓ | |
| … | |
| θα γράφω | ‘fut.ptcl write.1sg’ (by several steps involving sound change and analogy) |
The alternative path involved different steps and led to a different outcome; the starting point was the 3sg form (Table 6.23).
| θέλει γράφειν | ‘want.3sg write.inf’ |
| ⇓ | |
| θέλει γράφει | ‘will.3sg write.inf’ (by sound change deleting -ν#) |
| ⇓ | |
| θέλει γράφει | ‘will.3sg write.3sg’ (by reanalysis based on |
| || | convergence in form of 3sg and infinitive) |
| ||==⟹θέλω γράφω | ‘will.1sg write.1sg’ (doubly marked agreement |
| … | allowed by reanalysis) |
| ⇓ | |
| θε γράφω | ‘fut.ptcl write.1sg’ (by several steps involving sound change and analogy, as in Table 6.22) |
These two paths of development meant that Greek had both a future based on ‘want’ with the dms να and one without it, both arising via perfectly ordinary system-internal well-motivated processes of language change.Footnote 271 These WANT-based futures occur with great frequency in Medieval and Early Modern Greek texts from the thirteenth to the eighteenth centuries, and all the variant forms indicated in the above tables are attested, some even co-occurring in the same text. The types θε να and θα να, the predecessors to θα, are reported by Thumb 1912: §225) as present “dialectally or archaically” in the late nineteenth century, while Asenova 2002: 214 claims that θε να was limited to the fourteenth to eighteenth centuries and θα να was limited to the sixteenth century.
Greek is the only Balkan language in which no vestige of the subjunctive marker per se survives into the present day in future or future-like constructions, although the vowel of θα is testimony to the previous presence of να.Footnote 272 The fact that there were variants with and without να, i.e., θε να γράφω ~ θε γράφω, despite their different origins, may well have played a role in the partial absence of the dms të in Albanian, through a cross-language analogy (see §6.2.4.1.4), although the same development (elimination of the dms) in Balkan Slavic may also have been relevant.
Greek is also the only Balkan language in which ‘have’ (or the functional equivalent in Romani and WRT) plays no role in a modern future formation, though, as noted above, ‘have’ (ἔχω) was the basis for a future-type historically within Greek, and Asenova 2002: 217 gives historical variants of the ἔχω future with an inflected main verb and with an invariant (3sg) form of ἔχω, e.g., ἔχω να γράψω and ἔχει να γράψω. Thumb 1912: §226 does note the occurrence of ἔχω with a να-clause for the future in nineteenth-century Bova Greek of southern Italy (éh’yi na erti ‘he will come’) but that is not usual even now, where instead the use of the present for future reference is more common. That distribution, however, in a peripheral non-Balkan dialect, helps to localize the ‘want’ future as a Balkan feature, as does the absence of the ‘want’ future in Pontic Greek, where the modal marker να alone with an inflected verb can serve as the future (Drettas 1997: 298–304).
Further with regard to “non-Balkan Greek,” some facts about the Cypriot Greek future are telling in two respects. First, the [en] that figures in the Cypriot future tense is in some accounts from the past of ‘want,’ ἔθελεν (=> ἔθεν => ἔν) not from the present. Elsewhere in Greek, the past of ‘want’ is associated with the conditional (in the sense of a “future in the past,” see §6.2.4.2 and §6.2.4.2.4). Moreover, there is a future formation (Aerts 1983) with an invariant element με as the future marker, with the form με να γράψω ‘I will write,’ where the με is a reduced form of μέλλει ‘is about to,’ one of the other Postclassical future auxiliary competitors; given the phonic parallelism of θέλει and μέλλει, their parallel reduction reinforces the plausibility of the sets of developments each must have undergone.Footnote 273
These two facts, then, as with the Southern Italy and Pontic data, serve to localize the present of ‘want’ as the Balkan future prototype. Its relation to other WANT-based futures in the Balkans is explored in the various sections within §6.2.4.1.
6.2.4.1.2 Romance
Analytic futures begin to appear in Latin at the end of antiquity, using habeo ‘have,’ uolo ‘want,’ or debeo ‘have to’ + infinitive, or esse ‘be’ + future participle. In the rest of Romance the future ended up using forms descended from habeo (except in Logudoro (north-central) Sardinian, where the descendant of debeo became the marker), but in Balkan Romance, the uolo future became generalized. Nonetheless, Early Modern Romanian shows the competition in auxiliary choice, and the variation continues into the modern language (see Maiden et al. 2021: 362ff. et passim). For instance, regarding ‘have’ types, where Early Modern Romanian had ‘have’ (1sg am) + infinitive, e.g., am a bea ‘I will drink’ and ‘have’ (1sg am) + să + subjunctive, e.g., am să caut ‘I will search,’ modern Romanian has just the latter, though still with both necessitative and future meanings in the colloquial register (Zafiu 2013a: 38–39). Further, the Early Modern Romanian type with conjugated ‘want’ future marker (1sg voi(u)) + infinitive is still found in contemporary usage, e.g., vor vedea ‘they will see,’ while the type of conjugated ‘want’ with să + subjunctive has given rise to a type with invariant future marker o, e.g., o să merg ‘I will go.’ This latter type is colloquial, more typically southern, and dates to the seventeenth century. There is also a regional phonologically reduced conjugated future marker, e.g., 1sg oi (Zafiu ibid.).
In Aromanian, the marker is invariant (va, u, etc.) ±s’ + subjunctive, while in Meglenoromanian, as noted above, the future marker has been absorbed by the subjunctive marker and lost, except in Tsărnarekă where the marker ăs is used with the subjunctive. (See §6.2.5.7 on Meglenoromanian va + inf.) Aromanian dialects that are in contact with Macedonian also use the negative of ‘have’ for negative futures as a calque on Macedonian, e.g., Kruševo noare s’ neadzim / Belã di Suprã nori s’nedzim ‘we won’t go’ (Gołąb 1984a; Markovikj 2007).
It is also worth noting that, unlike Greek, Macedonian, SDBR, Albanian, or Romani, but like archaic or dialectal Bulgarian (using the extremely marginal short infinitive, see §6.2.4.1.3) and BCMS (regularly), the conjugated Romanian future marker can also be postposed to the main verb, which, as in the relevant Slavic languages, is an infinitive (Rosetti et al. 1969: 85–90, 267–268; Graur et al. 1966: 269–270).
6.2.4.1.3 Slavic
Vaillant 1966: 104–107 notes parallels among Baltic, Slavic, and Germanic expressions of futurity, e.g., by means of preverbs on presents that function as (future) resultatives (cf. also Večerka 1993: 184–185). In OCS, the future was usually expressed by a perfective present (imperfective for gnomic futures, Večerka 1993: 182) although there are periphrases using various auxiliaries plus the infinitive. The use of iměti ‘have’ parallels Greek and Latin as well as Gothic (haban), but only occurs as a true future (as opposed to obligative) in Bulgarian and Macedonian OCS texts, and, along with the Gothic, is likely a calque on the Greco-Roman usage. The use of na-/vŭ-čęti ‘begin.inf’ parallels the use of Gothic (but not West Germanic) duginnan ‘begin’ and is rare in OCS. The perfective present of ‘be,’ 1sg bǫdǫ, was used regularly as a future for that verb and in a variety of future periphrastic constructions (e.g., with past passive and resultative participles). In North Slavic as well as Slovene and the Kajkavian dialects of Croatian, the reflexes of bǫdǫ become the future auxiliary, with perfective presents also fulfilling future functions.Footnote 274 The future using ‘want,’ inf xŭtěti / 3sg.prs xoštětŭ, + infinitive is attested in OCS, albeit usually as a voluntative, and is regular as a future in South Slavic texts by the fourteenth century. The shortened form of the verb without the initial syllable appears in the thirteenth century, and the short infinitive in the sixteenth century. Conjugated short auxiliary + finite da-clause appears in the fifteenth century, and the invariant particle + da-clause appears in the sixteenth, although invariant particle + finite form, without dms da, shows up in the fifteenth century (Asenova 2002: 214).
For Bulgarian, constructions of the type particle + da + finite-verb, conjugated.auxiliary + da + finite-verb, and short infinitive + conjugated.auxiliary (e.g., šte da otideš, šteš da otideš, ‘you will go,’ napravi šteš ‘you will do,’ Stojanov 1983: 340, 386) survived as marginal into the nineteenth and early twentieth centuries and also in some dialects, but all are currently obsolete in the standard language. Moreover, in Bulgarian the particle šte is homonymous with the 3sg of the verb šta ‘want, like’ which continues to exist as a lexical verb. In Macedonian, kje + da has the effect of rendering the future suppositional, and there is no affirmative verb cognate with kje, but the negative nejkjam ‘I don’t want/like’ does continue the old finite verb (and see §7.4.1.2.2.1 and §6.2.1.1.6 on some variant forms of the ‘want’-based affirmative future marker in Balkan Slavic). The Torlak dialects of BCMS show a continuum from conjugated ‘want’ + da-clause (the infinitive is absent), to an invariant particle (če ~ će) + finite form as in Macedonian and Bulgarian.Footnote 275 In Leskovac, the 3sg clitic form of ‘want’ (će) is on the way to becoming the generalized particle. It can occur with all persons except the first (which uses ću), with finite forms, with or without da, and in sentence-initial position (Mihajlović 1977: 51). In the northeastern corner of the Timok-Lužnica dialect of Vratarnica (Sobolev 1994: 379), the situation is similar, although če is also attested with the first person (see also A. Belić 1905: 636ff.; Vukadinović 1996: 222–223; Remetić 1996: 502–503; Toma 1998: 278–279 for additional details).
The negative future shows a continuation of the use of ‘have’: the normal negated future in Balkan Slavic is the impersonal Mac nema, Blg njama (lit., ‘it hasn’t,’ but also ‘it does not exist’) + da-clause. The negation of the particle derived from ‘want’ can also occur, but often has voluntative overtones, just as the positive of existential ‘have,’ ima + da-clause, is obligative. In any case, the temporal, social, and geographic evidence argues strongly for the specifically Balkan nature of the Balkan Slavic future.Footnote 276 The place of the ‘have’ future has not been adequately appreciated in Balkan linguistic work, although Asenova 2002: 217–218 rightly turns attention to it. It is not simply the rise of the ‘want’ future but also the competition of the ‘have’ future that makes Balkan Slavic Balkan in this regard.
6.2.4.1.4 Albanian
Most general descriptions of Albanian identify as Geg the future using conjugated present of ‘have’ (1sg.prs kam in the standard and many dialects) + infinitive (= me + short participle in Geg, për të + participle [= për + nominalized participle] in Tosk) and as Tosk the future using an invariant particle derived from ‘want’ (do in the standard language and most dialects) ± të + subjunctive), the latter being typically Balkan, the former being identified as more similar to Western Romance. The actual distribution, however, is more complex and indicates the Balkan nature of the constructions. Moreover, ‘want’ + infinitive and ‘have’ + subjunctive also occur.Footnote 277 The ‘want’ + subjunctive construction is attested throughout Albanian from Baćica (Alb Baçica) in the Serbian Sandžak to Mouzakéïka (Alb Muzhakat) in Epirus (Alb Çamëri) as well as in diaspora dialects of Arbëresh in Italy, Arvanitika in southern Greece, and Arbanasi (Alb Arbëresh) in Croatia. Constructions with ‘have’ + infinitive occur mostly in central and northern Geg (including Arbanasi), usually in competition with the ‘want’ + subjunctive future (Gjinari et al. 2007: 376, Map 305). Both futures also occur in the oldest Albanian texts (Fiedler 2004: 531–532, 591ff.). Dialects with only the ‘have’ + infinitive future are rare and located at the northeastern and northwestern peripheries, where, however, nearby points have only the ‘want’ + subjunctive future. ‘Have’ + subjunctive futures are limited to Arbëresh in Calabria and Sicily. ‘Want’ + infinitive futures are found in the north bordering with the Zeta-Lovćen dialects of Montenegrin, and there the auxiliary can be conjugated, which makes the construction exactly analogous to standard BCMS, as in example (6.89) from Kelmend (Shkurtaj 1975: 55):
(6.89)
Jam i lik e duo me dek I.am pc.m.nom bad.m and want.1sg inf die.ptcp ‘I am ill and will die’
The ‘have’ + Tosk infinitive future is limited to a few scattered points in Northern Tosk, Lab, and Çam, and always co-occurs with the ‘want’ + subjunctive future. It is only the ‘want’ + subjunctive future that presents consistent areas where other futures do not compete. This area includes all of Tosk (with the isolated exceptions just noted), most of the Transitional and Southern Geg dialects, as well as most of East Central Geg and scattered points in the remaining Geg dialects (West Central, Northwest, and Northeast – with a compact area in northwestern Kosovo; see Gjinari 2007: 376, 390, and Friedman 2005b for details). For the various dialects where the two futures compete, in some one or the other predominates, while in others there is a pragmatic division, e.g., with ‘want’ functioning as suppositional or voluntative, ‘have’ as more obligational, etc. However, in Sh. Gjeçov’s Kanuni i Lekë Dukagjinit, representing traditional Northern Geg, ka + të + participle is used for permitted actions while do + të + subjunctive is used for obligations, as seen in example (6.90):Footnote 278
(6.90)
Dorëraras-i e ká të luejt-men naten e aty, kû t ‘a çilë drit-a, murderer-the it has pc move-ptcp at.night and he when dms him opens light-the do të struket. will dms hide.3sg.prs.mdp ‘The murderer [in a blood feud] may move around at night, but at the first light of dawn he must conceal himself.’ (Gjeçov 1989: 163–164 (Ch. 119, §849))
One noteworthy feature of the WANT-based future in Albanian is that the dms të is optional, so that both do të takoj and do takoj are acceptable, for ‘I will meet,’ with the latter being more colloquial. It is possible that the absence of të is merely the result of an allegro reductive process, since an unstressed ë is particularly susceptible to elision in fast speech. That is, do të VERB could have been elided to do t VERB, and if the main verb began with a consonant, cluster reduction could have led to the ultimate effacing of të, e.g., do të takoj ‘I will meet’ => do t takoj => do takoj.Footnote 279 However, in such a scenario, we might have expected to see do t persisting before a vowel (e.g., do t emëroj ‘I will name’) or before phonologically congenial consonants such as sibilants (e.g., do t sugjeroj ‘I will suggest’), given that Albanian has affricates that are phonetically quite similar to combinations, i.e., [t] + [s/∫] would show strong similarity to Albanian < c >/< ç >, respectively. No such phonologically conditioned variation between do and do t seems to be evident, however.
Thus there may have been another mechanism at work in the do ~ do t(ë) variation. In particular, as noted in §6.2.4.1.1, Greek, through language-internal processes, developed both a future with a particle alone (θε γράφω ‘I will write’) and one with a particle plus a dms (θε να γράφω ‘I will write’). This variation within Greek of θε να γράφω ~ θε γράφω could have influenced the Albanian do ~ do t(ë) variation via contact, if Albanians – particularly those in Tosk territory, where knowledge of Greek is and was widespread – modeled their future on these Greek forms, as a sort of cross-language proportional analogy, what is essentially a calquing mechanism:Footnote 280
(6.91)
θε να γράφω : θε γράφω :: do të shkruaj : X, X => do shkruaj
In this account, then, both the development of the do të VERB future and the emergence of the do VERB future could have been affected by language contact.Footnote 281
6.2.4.1.5 Romani
As noted in §6.2.4.1 above, the Romani dialects of the Balkans, including dialects of Balkan origin such as Crimean, form the future with a particle derived from a verb meaning ‘want, like, love,’ usually ka[m/n]-< kam- (in Drindari (eastern Bulgaria) mə < mang- ‘idem’). Given the absence of such a future in Romani dialects outside the Balkans, this is clearly a Balkanism in Romani (Boretzky & Igla 2004: 1.244).Footnote 282
In this connection we address here the question of a morphological subjunctive in Romani. According to Matras 2002: 155, the original pattern in Romani was that the present (which also functioned as a future) suffixed -a to the person marker and the subjunctive dropped that ending, e.g., 3sg indicative kerela, 3sg subjunctive kerel ‘do’ (but see now Scala 2022). In many of the dialects that left the Balkans at roughly the time of the Ottoman conquests, the long present in -a developed into the future (Boretzky & Igla 2004: 1. Map 138, 2.172–174). The dialects that remained in the Balkans developed the ‘want’ future, as in the other Balkan languages, and there was an intermediate stage during which the subjunctive marker followed the future marker (ka te + finite verb), a construction that still occurs in Romani, albeit infrequently in most dialects (cf. the use of kje da in Macedonian §6.2.4.1.3). In these dialects, i.e., those of the Balkans, there is a general tendency for the Ø-ending to be used after te and ka while the so-called long form in -a is used in other present tense contexts. In VAF’s field notes and recordings from North Macedonia, this is true in the majority of cases, but not all. On occasion, long forms occur after ka and te and short forms occur in main clauses (pace Matras 2002: 156; see Friedman 2024). The same type of variation is also attested in the Bugurdži dialect of Kosovo (Boretzky 1993: 177, 187), and in various Arli dialects (Cech et al. 2009: 168 et passim; Cech & Heinschink 2002). At least in the case of main clauses, it appears that the alternation of long and short forms serves as a narrative device expressing focus or emphasis, and there is a geographic tendency for short forms to dominate as one moves further north (Friedman 2018e, 2024). The precise distribution of this discourse phenomenon requires further study. The point here is that while the Romani of the Balkans resembles Albanian and Balkan Romance in having a quasi-distinct subjunctive (which, in those latter languages, is also used after the future marker in dialects that lose the subjunctive marker in such constructions), nonetheless, the variation is such that the distinction cannot be taken as absolute.Footnote 283
Moreover, those dialects in contact with Balkan Slavic sometimes calque the negative ‘have’ future using a possessive construction. In Romani, as in the rest of Indic and most of Asia, and including Russian and Finnish but excluding East and Southeast Asia and also greater Iran (Masica 1976: 166–169), the concept of ‘have’ is expressed analytically, in the case of Romani by ‘be’ + accusative. Example (6.92a) is a negative future calqued on the Balkan Slavic model, while (6.92b) is a calque on a positive that could be construed as obligative, but in Prizren Arli is the ordinary future, probably as a result of Albanian influence.
a.
Nae man te džav not.is me.acc dms go.1sg.prs ‘I won’t go.’ b.
si man te džav is me.acc dms go.1sg.prs ‘I have to go’ ~ ‘I will go’
Dialects of Romani with extensive Turkish conjugation have three options for forming the future with Turkish verbs: Romani future marker + Turkish present, Romani future marker + Turkish optative as in (6.93), and Turkish future as in (6.94) (from Friedman 2013b, cf. Table 6.18 and Footnote footnote 200 above):
(6.93)
Kidal kam diištir-elim e dasengo dišinmenki e romenge askal thus fut change-opt.1pl the Bulgarian.pl.gen thinking the Rom.pl.dat about ‘Thus we will change Bulgarian thinking about Roms’ (Futadži, Ivanov 2000)
(6.94)
Ame naši dön-dže-s ži kana doorul-ma-jə odia we can’t return-fut-1pl until when get.well-neg-prs that.f ‘We cannot go back, until she gets well.’ (VG 384, RMS)
The use of the future marker with an optative appears to be a calque on the older form of the Balkan future, which uses an analytic subjunctive clause.
6.2.4.1.6 West Rumelian Turkish
Finally, the West Rumelian dialects of Turkish, which like Romani use an analytic construction to express ‘have’ (positive existential var, negative existential yok), calque the Balkan Slavic negative future (see §6.2.1.4.3) using an optative to translate the da-clause (Friedman 1982c) (6.95):
(6.95)
yok-tur gidelim negative-existential.copula go.1pl.opt ‘we won’t go.’
6.2.4.1.7 Judezmo
In the case of Balkan Judezmo, the crucial datum is the favoring of analytic over synthetic constructions. Like Modern Spanish (and English, French, etc.), Judezmo can use a verb meaning ‘go’ to mark futurity, although it also has at its disposal the non-Balkan Romance synthetic future, itself derived from (Late Latin) infinitive + ‘have.’ However, Kramer & Perez-Leroux 2007, based on a ten-page text in Crews 1935, observe that out of forty futures only two were synthetic, and those were both in more formal contexts. On the other hand, the analytic ‘go’ future is common everywhere in colloquial Spanish, especially in Latin America, where the synthetic future is increasingly rare. The fact that Latin America is the other place where the synthetic future is most rare could be significant, since the timing of the separation of Latin American Spanish coincides roughly with the separation of Judezmo. One could even speculate that the two contact environments each favored such a development. On the other hand, it could simply be parallel continuations of internal drift. Nonetheless, based on various studies of Latin American Spanish (e.g., Orozco 2007 and the literature cited therein), it appears that Judezmo has gone significantly further than any Spanish dialect in this regard. In Continental Standard Spanish, the ‘go’ future is more frequent colloquially, but the synthetic future is vastly more common in written texts. Moreover, the two futures are not entirely interchangeable in Standard Spanish.Footnote 284 Although it requires further study, it is possible that the various Balkan analytic futures influenced the degree to which the analytic replaced the synthetic future in Judezmo.
6.2.4.1.8 Futures: Summary
Drawing on the discussion in the preceding sections, the parameters for variation in the future tense in relevant languages of the Balkans can be summarized as follows for both WANT-based and HAVE-based futures (see Table 6.24).Footnote 285
These parameters and the forms that they determine are summed up in Table 6.25, with an indication of the parameters along with ‘want’ versus ‘have’ as the relevant auxiliary verb; since some languages allow both settings for a given parameter, there is a wide range of variants. Note that “infl.aux” refers to parameter (b), “dms” to parameter (c), and “inf” to parameter (d), all treated as binary settings of + or -, except for most of Meglenoromanian, where Ø marks the unique situation where the fut element is deleted in favor of the dms. The Roman numerals with “Language” indicate earliest century of attestation when available, which, in the case of all the languages except Greek and Balkan Slavic, is limited by the lateness of or gaps in (for Balkan Romance) the documentation.
| LANGUAGE | WANT/HAVE | infl. aux | dms | inf | EXAMPLE |
|---|---|---|---|---|---|
| Greek (I–XI) | HAVE | + | − | + | ἔχω γράψειν |
| Greek (XII) | HAVE | + | +Footnote * | − | ἔχω να γράψω |
| Greek (XIV) | HAVE | − | + | − | ἔχει να γράψω |
| Greek (VII) | WANT | + | − | + | θέλω γράψειν |
| Greek (XIV) | WANT | + | + | − | θέλω να γράψω |
| Greek (XV) | WANT | − | + | − | θέλει να γράψω |
| Greek (XV) | WANT | − | − | − | θε [να] γράψω |
| Greek (XVI) | WANT | − | − | − | θα να γράψω |
| Greek (XVI) | WANT | − | − | − | θα γράψω |
| Latin (II) | HAVE | + | − | + | habeō cantare |
| Latin (V) | WANT | + | − | + | volo cantare |
| Romanian (XVI–XVII) | WANT | + | + | − | voi(u) să scriu |
| Romanian (XVI) | WANT | + | − | + | voi scrie |
| Romanian (XVII/XVIII) | WANT | − | + | − | o să scriu |
| Romanian (XVII) | HAVE | + | + | − | am să scriu |
| Romanian (XVI–XVII) | HAVE | + | −Footnote 286 | + | am a scrie |
| Meglenoromanian | WANT | Ø | + | − | si fac |
| Meglenoromanian | WANT | − | + | − | ãs fac |
| Meglenoromanian | WANT | − | − | + | va veári |
| Meglenoromanian | HAVE | + | + | − | am si fac |
| Aromanian | WANT | − | + | − | va s-cântu |
| Aromanian (some) | WANT | − | − | − | va cântu |
| Aromanian (some) | HAVE | − | + | − | noare s’ neadzim |
| OCS (X–XIV) | WANT | + | − | + | xoštǫ pisati |
| Church Slavonic (XIII–XV) | WANT | + | − | + | štǫ pisati |
| Church Slavonic (XVI–/Blg XX) | WANT | + | − | + | šteš pozna/pozna šteš |
| Balkan Slavic (XV–XX) | WANT | + | + | − | šteš/kješ da imaš |
| Balkan Slavic (XVI–XX [Blg]/XXI[Mac]) | WANT | − | + | − | šte/kje da imaš |
| Balkan Slavic (XV) | WANT | − | − | − | šte/kje imam |
| OCS (X–XIII/XIV) | HAVE | + | − | + | imamь pisati |
| Balkan Slavic (XVIII) | HAVE | − | + | − | njama/nema daimam |
| Balkan Slavic(XVIII) | HAVE | + | + | − | imam da imam |
| Torlak (BCMS) | WANT | + | + | − | ću/ču [da] idu/idem |
| Torlak (BCMS) | WANT | − | + | − | će/če [da] idu/idem |
| Torlak (BCMS) | WANT | + | − | − | dogovoríču/ću |
| Albanian (XVI) | WANT | + | + | − | dua të shkruaj |
| Albanian (XVI)Footnote 287 | WANT | − | + | − | do të shkruaj |
| Albanian (XVIII) | WANT | − | − | − | do shkruaj |
| Albanian (Geg) | WANT | + | − | + | duo me shkrue |
| Albanian (Tosk) | HAVE | + | + | − | kam për të shkruar |
| Albanian (Geg) (XVI)Footnote 288 | HAVE | + | − | + | kam me shkrue |
| Albanian (Arbëresh) (XVI) | HAVE | + | + | − | kam të shkruaj |
| Albanian (Arbëresh) (XX) | HAVE | − | + | − | ka të shkruaj |
| Romani | WANT | + | + | − | ka te džav |
| Romani | WANT | − | − | − | ka džav |
| Romani | HAVE | − | + | − | nae man te džav |
| West Rumelian Turkish | HAVE | − | − | − | yoktur gidelim |
* A very few examples of έχω + finite verb without the dms να occur in Medieval Greek but there is reason to consider each one to be an error; see Holton et al. 2019: 1780 for examples and discussion
6.2.4.2 Conditionals: An OverviewFootnote 289
The Balkan languages show numerous striking convergences, but also some differences, with regard to conditionals, defined here both in terms of their form – especially in the blend of future marking and past marking – and in terms of their function, having to do with their use in expressions of modality and in ‘if … then’ clausal combinations. The complete or near-complete agreement of the expression of irreal (counterfactual) conditionals (nonexpectative past in Hacking’s 1997a terms, expectative unfulfillable in Kramer’s 1986 terms) among Greek, standard Macedonian and the dialects on which it is based, Aromanian, and Tosk Albanian was first noticed by Sandfeld 1930: 105, and the complete elaboration of the historical developments of Balkan Slavic, Balkan Romance, and Albanian together with Greek was achieved in Gołąb 1964a. Kramer 1986 examines the full set of conditionals in Macedonian, Hacking 1997a builds on Kramer’s work with her comparison of Macedonian and Russian, and Belyavski-Frank 2003 completes the dialectal and discourse pragmatic picture for the four main Balkan languages/groups as well as the non-Torlak dialects of BCMS (cf. also Asenova 2002: 220–239 and Cugno 1996). Montoliu & van der Auwera 2004 discuss Judezmo in this regard, with comparisons with other Balkan languages. Moreover, the Balkan character of conditionals in the Romani dialects of the Balkans has been studied by Friedman 2014b. The main point of relevance here is that the intersection of future and preterite (usually imperfect, but sometimes also perfect and pluperfect) marking (both perfective and imperfective in those languages with the distinction) came to mark irreal/counterfactual conditional and iterative-habituals as well as anterior futures. The intersection of these categories is not by itself a peculiarity of the Balkans (cf. Aronson 1977), but the development of how they are marked (and, as Belyavski-Frank has shown, the distribution of related pragmatic functions) is distributed and attested in such a way that we can speak of a Balkan conditional as a Balkanism, i.e., an areal and not just a typological feature.
The Balkan conditional (Gołąb 1964a; Belyavski-Frank 2003) is formed by the intersection of future and past markers, i.e., the anterior future becomes a conditional (cf. Sandfeld 1930: 105). As with the volitionally based future, this is a typological commonplace that nevertheless can be identified as a Balkanism when the historical facts are examined. The Balkan conditional presents a less uniform picture than the Balkan future, but the basic parameters are comparable, and the southwestern Balkans again emerge as the center of innovation. As with the future, Old Indic, Early Latin, and Ancient Greek all had synthetic modal formations that fill some of the functions relevant to a consideration of the Balkan conditional, and a synthetic type based on Latin occurs in some of Balkan Romance (see §6.2.4.2.2); Common Slavic, however, entered the Balkans with a dedicated analytic conditional in place.Footnote 290 The Balkan construction itself can have a variety of related meanings, e.g., classic irrealis (‘X would have happened but did not’), potential (‘X would happen if Y’ [hypothetical and expectative (Kramer 1986, see §6.4.2.1)], including ‘X almost happened/was about to happen’), hypothetical, iterative-habitual, anterior future, presumption, and attenuation, and languages and dialects can be differentiated on the basis of which of these meanings are encoded (see §6.2.4.2.1 for a detailed discussion of the framework used here).Footnote 291 Of interest here are the obvious convergences (Balkanisms) and the principal points of difference, and these are discussed here and in the sections that follow. See Table 6.26 for a language-by-language summary.
The most Balkan (or grammatically integrated) construction is analytic and consists of the same particle that marks futurity plus a past tense form (imperfect, perfect, or pluperfect). This construction is characteristic of most of Macedonian, of Bulgarian dialects in the southeastern Rhodopes (including Pomak in Greece, Kokkas 2004: 174) and west of Kjustendil, of colloquial Tosk Albanian, of all of Greek, of southern Aromanian, and of Romani in the Balkans. Slightly less grammatically integrated (and older) is the future marker plus dms plus past tense, which is found in the Albanian of the Tosk-based standard, and is also the northern Aromanian construction. This construction also occurs in dialectal western Macedonian (Gołąb 1964a: 47; Belyavski-Frank 2003: 161; Koneski 1981: 173), as well as in northwestern Bulgarian (e.g., Vidin region, M. Mladenov 1969: 105), and in southeastern Macedonian (e.g., Ser (Grk Sérres), Asenova 2002: 237), etc., where, as in Tosk, it sometimes coexists with the construction without the dms, the difference being that the dms is prescribed by the Albanian standard, but not by the Macedonian one. However, the Macedonian can be used as a suppositional. Meglenoromanian, which has merged the future and present subjunctive (except in Tsărnarekă), nonetheless uses invariant vrḙa ‘want/will’ plus dms plus the present and perfect to form conditionals (Atanasov 1990: 226, pace the older sources cited in Belyavski-Frank 2003: 245–246). Next down on the scale of greater grammatical integration (or less so, and thus more archaic) is invariant imperfect marking on the verbal particle (historically, the 3sg) plus dms plus nonpast, found in Torlak BCMS, e.g., teše da idu ‘I would have gone,’ adjacent Macedonian (Kumanovo-Kriva Palanka), and Bulgarian (western transitional) dialects.Footnote 292 Still less grammatically integrated is the conjugated imperfect of the verb, which is also the source of the future particle, together with a nonpast, usually with the dms. It is characteristic of most Bulgarian dialects and the standard language and also Balkan Romance. Romanian also has a future in the past with the imperfect of ‘have’ (cf. Bara et al. 2005:180).Footnote 293 At the farthest periphery, the imperfect of ‘want’ plus infinitive in conditional-type meanings occurs in the South Slavic of Bosnia-Hercegovina, Montenegro, and Kosovo as well as in Banija, Kordun, Lika, and coastal Croatia south of there, western Serbia and Srem (Belyavski-Frank 2003: 18, 272–274). This is a continuation of the oldest attested construction. Remnants of this construction also occur in some peripheral Macedonian (Ser (Grk Sérres), Drama) and Bulgarian (Rhodopian) dialects using a short infinitive (Asenova 2002: 237). Those Geg Albanian dialects that use the ‘have’ + infinitive future employ an analogous conditional, namely a conjugated imperfect of ‘have’ with the infinitive, as did Postclassical Greek, though by Medieval Greek (Holton et al. 2019: 1795–1814), a variety of similar ‘want’-based formations, most notably and most predominantly imperfect plus an infinitive, arose, parallel to the innovative ‘want’-based futures (see §6.2.4.1).Footnote 294 Moreover, in those languages and dialects where negated existential ‘have’ marks negated futurity, the corresponding negative imperfect can form a conditional. Most of Romanian uses a special conjugated conditional auxiliary (aş, ai, ar, etc.), whose origin is a matter of some debate (either ‘have’ or ‘want’; see §6.2.4.2.2), plus the bare infinitive, a type that was in wide use in Early Modern Romanian,Footnote 295 but is not found in all dialects of Romanian,Footnote 296 nor in Aromanian nor in Meglenoromanian (see §6.2.4.2.2 for details). Romanian can also form past conditionals with the conditional marker plus fi ‘be,’ plus past participle, a construction which is also a past presumptive (cf. §6.2.5.7; Zafiu 2013a: 50).
The sequence of changes that gives the various forms is evident for Slavic and Greek (as with the future proper, the initial stage is an independent development in each language that was in competition with other possibilities) and can be presumed for the others. It has three principal stages (see Figure 6.1): for the ‘want’-based forms (development I), the imperfect ‘want’ + infinitive (1) developed into (2) imperfect ‘want’ + dms + present, via the infinitive-replacement process that was on-going in each language (cf. §7.7.2.1); that combination became (3) future/modal marker ±dms + imperfect. There is, however, another set of stages that is attested in Macedonian dialects, and is among the variants in Medieval Greek, which we can label (2º), namely imperfect ‘want’ ± dms + imperfect, e.g., Mac kješe [da] dojdeše ‘s/he would have come’ (Koneski 1981: 173) or MedGrk ήθελεν έπαιρνε ‘s/he would have taken’ (Holton et al. 2019: 1811). Stage (2º) could be viewed as an extension of imperfect marking to the main verb, which was subsequently lost on ‘want’ (stage 3) in much the same way that the conjugated present of ‘want’ became an invariant future marker. In this scenario, the marking for imperfect did not flip (see Footnote footnote 294), but rather leaked (see also §6.2.4.2.4, near the end). Given that the infinitive replacement is generally dms + present, the question arises concerning whether (2º) was an intermediary stage between (1) and (3) or between (2) and (3), i.e., how early did the leakage occur? There are three possible scenarios: (a) there were two separate paths from (1) to (3), one via (2) and the other via (2º); (b) stage (2º) came between (1) and (2); (c) stage (2º) came after (2), a “detour” as it were, on the way to (3). Are we dealing with multiple scenarios involving different regions? The paucity of textual evidence does not permit a definitive answer for Balkan Slavic, and although there are some (2°) examples for Greek (see immediately above), suggesting the “detour” scenario (c), they are few so that again a definitive answer is not easily arrived at. The final reanalysis in (3), like the changes in clitic order (cf. §5.5.2, §5.5.3, and §7.3.3), put Macedonian (and some Bulgarian dialects) with Greek and (Tosk) Albanian and (northern) Aromanian, while most of Bulgarian (and some Macedonian dialects) remain like Medieval Greek. Moreover, as with the future, ‘have’ constitutes a significant competitor with ‘want,’ in which case the stages for this development II are these: (1’) imperfect ‘have’ + infinitive developed into (2’) imperfect ‘have’ + dms + present, via the infinitive-replacement as with I.1-2. A somewhat different path is that in III (with III = I.1 or II.1’, with the development from I.1 (or II.1’) into (4) with a conditional auxiliary (via reduction and reanalysis of the WANT- (or HAVE-) form) + infinitive.
Schematically, these developments can be represented as in Figure 6.1, with different numbers used to indicate the different outcomes, and apostrophes used to indicate similarities among the types where appropriate. For stage (2º) the ambiguities are represented by a slash.
Of these types, Greek has I.3 (Medieval Greek has I.2), Macedonian I.3 for the positive and II.2’ for the negative, and Bulgarian I.2 for the positive and II.2’ for the negative; Tosk Albanian has I.3, as does Aromanian (although some Aromanian dialects also preserve the synthetic conditional); pre-World War Two Standard Geg Albanian has II.1’.Footnote 297 Romanian has III.4 primarily, but regionally I.1 is found, and II.2’ occurs as a future-in-the-past (see Footnote footnote 266); Meglenoromanian, for its part, has a variant of I.2 in which the 3sg.impf vrḙa functions as an invariant particle (Atanasov 2002: 251). Romani ka džalas~džala sine ‘he would go’ is also type I.3.Footnote 298 In sum, then, the Balkan conditional resembles the Balkan future both in its choices of auxiliary and in the relative degrees of expansion and development. In both categories, the nexus of Tosk Albanian, Aromanian, Greek, and Macedonian along with co-territorial Romani dialects emerge as the most convergent, and the basic type of innovation (in terms of auxiliary choice) extends into non-Torlak BCMS.
As to the timing of these developments, Zafiu 2013a: 62 notes that such uses of the conditional occur in Romance outside the Balkans and elsewhere. The point, however, is to examine the history of the feature in the Balkan languages rather than comparing modern synchronic states. In the case of Slavic, not only is Balkan Slavic unique within Slavic in developing such usages – in competition with an inherited conditional – but it is also the case that the development took place during the period of Balkan language contact. Similarly in Greek, while early instances of modal/conditional uses of a past of ἔχω ‘have’ with an infinitive occur in the Koine period, it is during the late Byzantine and Medieval Greek periods that conditional forms and usage become prevalent and more stable (Horrocks 2010: 130–131; 298–300).
The main types found for conditionals in the Balkan languages are summarized in Table 6.26, and the facts from the individual languages are surveyed in the subsequent sections, especially as to the functions these forms are put to.Footnote 299
| Romani | ka | keravasFootnote * | ‘I would have done’ | ||
| Greek | θα | έκαναFootnote ** | |||
| Macedonian | kje | napravevFootnote *** | |||
| Aromanian | va | [s] | fãceamuFootnote **** | ||
| Albanian (Tosk) | do | të | bëja | ||
| Meglenoromanian | vrḙa | si | am fat(ă) | ||
| Bulgarian | štjah | da | napravja | ||
| Torlak BCMS | teše | da | napravim | ||
| other BCMS | šćaše/šćeše/ | da | učinim | ||
| teše | |||||
| fut | dms | do.impf.1sg | |||
| Albanian (Geg) | [kishna | me | bâ] | ||
| I.have | infm | do.ptcp | |||
| Romanian | aş | fi | făcut | ||
| cond | be.inf | do.pst.ptcp | |||
| TurkishFootnote † | yap | acak | tı | m | |
| root | fut | pst | 1sg |
* Some Arli dialects have a new imperfect formed by the present + be.3sg.impf, e.g., kerava [s]ine
** Greek ‘do’ (κάνω in the standard language, κάμω in regional dialects) does not readily distinguish imperfect past from aorist past, so ‘write’ is shown here, as it does make that distinction formally (imperfect έγραφα versus aorist έγραψα). Other examples use ‘do.’
*** Dialectal Macedonian has inflected kješe + da + prs; Kumanovo and Aegean Macedonian can also have the full verb sakaše + da + prs, etc. (see Belyavski-Frank 2003: 268 for further details).
† Turkish has a quite different conditional system, as discussed in §6.2.4.2.6, but the example adduced here is the one that comes closest to the other examples.
6.2.4.2.1 Balkan Slavic
As observed in §6.2.4.2, Footnote footnote 290, Common Slavic (as evidenced by OCS) had analytic conditional constructions consisting of an old optative of ‘be’ (3sg bi) or the prfv.aor of ‘be’ (3sg by), or the prfv.prs of ‘be’ (3pl bǫdǫtŭ) plus the resultative participle. This last-mentioned formation functioned as an anterior future (Večerka 1993: 184–185), but the first two were for irrealis conditionals. In Slavic outside the Balkans, the inherited system was more or less retained, insofar as the conditional continues to be expressed by a descendant of ‘be’ plus the descendant of the resultative participle, although details vary from language to language. For Slavic, it is only in the Balkans that a new conditional, as described in the overview in §6.2.4.2, developed. At the same time, the inherited conditional is retained in all of Balkan Slavic, except in the southwestern Macedonian dialects, where the descendant of the resultative participle survives only in a few evidential/admirative usages (see §6.2.5.1), and therefore all the paradigms based on that participle have been replaced. Macedonian differs from Bulgarian (and Torlak BCMS) in that the old auxiliary is now an invariant particle (bi).Footnote 300 This combination of innovation and retention results in a complex set of realis/irrealis relations. Kramer 1986 identifies two pairs of oppositions that together yield four combinatory possibilities for Balkan Slavic conditionals. In the protasis (if-clause), the opposition is expectative/hypothetical, while in the apodosis (then-clause), the opposition is fulfillable/unfulfillable.Footnote 301 Hacking 1997a provides a somewhat different interpretation. She treats hypotheticality as the basic parameter of all conditionals, within which the basic opposition is expectative/nonexpectative, and within nonexpectative she identifies a three-way temporal opposition past-present-future, depending on the reference time of the statement. The advantage of both these systems over a traditional tripartite division into real, irreal (counterfactual), and potential is that they capture a three-way distinction within the nonreal group of conditionals. This means that in addition to counterfactual and potential there is an intermediary category that has a distinct realization in Balkan Slavic but not in English, namely the hypothetical unfulfillables (in Kramer’s terms) or the present nonexpectatives (in Hacking’s). These are discussed in connection with examples (6.99–6.101) below.
The four combinations arising from these two oppositions are illustrated for Macedonian (M) and Bulgarian (B) in (6.96)–(6.101), and then discussed below each pair.Footnote 302
Present Expectative (Fulfillable)
(6.96)
(M) Ako mi se javite kje dojdam. (B) Ako mi se obadete šte dojda. if me.dat intr call.2sg.pfv.prs fut come.1sg.pfv.prs ‘If you call me, I will come’ (i.e., definitely come)
Examples (6.96) are typical real conditionals. If the dms is substituted for ako, the degree of expectation is attenuated but not eliminated. The future in the apodosis puts the focus on intention. Past expectatives (in Hacking’s terms) use the fut + perfective imperfect in the main clause to denote past iterative or habitual actions.
Hypothetical Fulfillable (Potential, Future Nonexpectative)
(6.97)
(M) Ako mi se javite bi došol. (B) Ako mi se obadite bih došel. if me.dat intr call.2sg.pfv.prs cond come.pfv.lf.m ‘If you called/were to call me, I would come’ (with pleasure; albeit without enthusiasm; etc.)
In (6.97), the apodosis puts the focus on desire (whether positive or negative) rather than intention. Note that with the dms rather than ako, the interpretation of the sentence in Macedonian depends on the nature of the protasis. In the present example, substituting da for ako does not significantly change the meaning other than by adding the note of attenuation referred to above. If, however, the protasis sets an unfulfillable condition, i.e., if the conditional is hypothetical but unfulfillable (as discussed in example (6.100) below), Macedonian can still have a present tense in the protasis, but Bulgarian cannot.
Expectative Unfulfillable = Past Nonexpectative = Irreal
Expectative unfulfillables (Kramer) or past nonexpectatives (Hacking) are traditionally called irreal. For these conditionals, use of da rather than ako in the protasis is felt to be somehow old-fashioned or rural for Bulgarian (6.98B) but not for Macedonian (6.98M), while the use of a perfective imperfect rather than a pluperfect in the protasis is rejected for Bulgarian as sounding artificial. This latter is a major difference between Bulgarian and Macedonian. Although in principle a pluperfect could also be used in Macedonian, in fact the pluperfect is so marginal in the language that it is almost never met with in modal constructions, and even forms marking taxis have become quite rare (cf. §§6.2.3.2, 6.2.3.2.1, 6.2.3.2). The apodosis shows the expected difference between the encoding of ‘past’ on the main verb (Macedonian) versus the auxiliary (Bulgarian):
(6.98)
(M) Ako/Da mi se javevte, kje dojdev. if me.dat refl call.2pl.impf.pfv fut come.1sg.impf.pfv (B) Ako/Da mi se bjahte obadili, štjah da dojda. If me.dat refl be.2pl.impf call.lf.pl fut.1sg.impf.pfv dms come.1sg.prs.pfv ‘If you had called me, I would have come.’
Hypothetical Unfulfillables = Present Nonexpectatives
Hypothetical unfulfillables/Present Nonexpectatives express an unfulfillable condition in the present. Bulgarian permits the conditional-imperfect in both clauses of a complex conditional sentence, which seems more frequent in negatives, as in (6.99), with the Macedonian given for comparison:
(6.99)
(B) Amerika da bjaxme, njamaše da izdăržim … (M) Amerika da bevme, nemaše da izdrižime … America dms be.1pl.impf neg.FUT.PST dms endure.1pl.prs ‘If we were America, we wouldn’t put up with it … ’
As can be seen from (6.100 M/B), the two languages also differ in choices for the protasis of hypothetical unfulfillables; where Bulgarian uses an imperfect, Macedonian can use a present.
Hypothetical Unfulfillable
(6.100)
Footnote 303(M) Da može bebeto da prozboruva, bi ti reklo … dms can.3sg.prs baby.def dms speak.3sg.prs.ipv cond you.dat say.pfv.lf.n ‘If the baby could talk it would say to you … ’ (B) Da možeše bebeto da (pro)govori, bi ti kazalo… dms can.3sg.imp baby.def dms speak.3sg.prs.ipv cond you.dat say.pfv.lf.n ‘If the baby could talk it would say to you … ’
Bulgarian can also use the inherited conditional in hypothetical unfulfillables, and in Macedonian the inherited conditional with bi can also be used in the protasis, as in (6.101).
(6.101)
6.2.4.2.2 Balkan Romance
Balkan Romance conditionals have several elements that are familiar from a Balkan standpoint but also some details that are unique. The formal side is where the unique elements are to be found while the familiar tends to be on the functional side.
A general outline of the Balkan Romance conditional from a formal standpoint is given in §6.2.4.2, but as noted in Footnote footnote 266, there are numerous details to be added, as well as some historical considerations, all needed to complete the picture. One of the unique elements about conditionals in Balkan Romance is that even though the most prevalent Balkan type, an analytic formation based on elements paralleling the future tense, is found across the languages, in varying forms, nonetheless a synthetic conditional that continues a synthetic Latin form persists into the historical period and is reported in some accounts well into the twentieth century.
Thus, for Romanian, Zafiu 2013a: 41 observes that “In old Romanian, there existed also a synthetic conditional [e.g., cântare ‘I would sing’], inherited from Latin … [perhaps] from the Latin future perfect [e.g., cantāverō], from the perfect subjunctive [e.g., cantāverim] … or from their contamination”; she writes further (Zafiu 2016: 23) that the synthetic form is found in sixteenth-century texts, “employed in the protasis of conditional constructions, commonly interpreted as a present conditional, but similar to a future.” Such a synthetic form, however, does not occur in present-day Romanian.
For Aromanian, Capidan 1932: 470–473 (and sources cited therein), Caragiu-Marioțeanu 1968: 106, 125–126, Papahagi 1974: 66–67, Saramandu 1984: 458–459, and Vrabie 2000: 63–64 give a synthetic form, preceded by the dms s(i), for a present conditional, e.g., 1sg s-cântárim, 2sg s-cântárişi, 3sg s-cântári, etc. ‘were I/you/(s)he to sing.’Footnote 304 Capidan 1932: 471 notes that the synthetic conditional is used in the south but is very rare in the north. Ianachieschi-Vlahu 1993: 85, 2001: 128 gives the synthetic conditional but writes that such forms do not exist in many Aromanian dialects, and Gołąb 1984a: 129–130, Bara et al. 2005: 195–200, and Markovikj 2007 do not give them at all but cite only analytic formations with the fut marker va (or a variant thereof) ±dms with the imperfect of the main verb. Gołąb 1984a: 107 states explicitly that the synthetic conditional does not occur in Kruševo. Caragiu-Marioțeanu 1968: 111–112, Papahagi 1974: 66–67, Saramandu 1984: 459, and Ianachieschi-Vlahu 1993: 84–85 – but not Ianachieschi-Vlahu 2001: 128 – and Vrabie 2000: 63–64 also give forms for a synthetic past conditional.Footnote 305 The analytic imperfect conditional consists of invariant vrea (3sg imperfect of ‘want’) plus dms plus the imperfect, aorist or synthetic present conditional, or the conjugated imperfect subjunctive of ‘have’ + the present conditional (Caragiu-Marioțeanu 1968: 112).Footnote 306
And, with regard to Meglenoromanian, Capidan 1932: 471–473 makes it clear that the synthetic type does not occur. Thus, overall, the evidence in Balkan Romance points to a continuation of the Latin synthetic conditional surviving into the last century, but being moribund at best, losing ground in the face of competition from a (more Balkan-like) analytic formation.
As interesting as these historically attested synthetic forms are, the most noteworthy fact about conditionals in Balkan Romance is the occurrence of a construction in Romanian, and only Romanian, with a dedicated auxiliary governing an infinitive. Such a type is found nowhere else in the Balkans, and importantly, occurs neither in Aromanian nor in Meglenoromanian. Thus, Romanian forms its conditional analytically with the short infinitive preceded by a special auxiliary, the conjugation of which is given in (6.102):
(6.102)
Romanian Conditional Auxiliary 1sg aş 1pl am 2 ai 2 aţi 3 ar 3 ar (Early Modern Romanian: ară/are for both 3sg and 3pl)
Some forms in this paradigm are identical with the auxiliary used with the past participle to form the perfect: the 2sg, 1pl, and 2pl. However, the other forms, 1sg, 3sg, and 3pl, are different so that the overall paradigm can be considered to be a specialized set for the conditional. Thus in Romanian we have aş face, ‘I would do,’ ai face ‘you would do,’ etc. As noted in §6.2.4.2, this type is in widespread use already in the oldest Romanian texts, those from the sixteenth century, though with less cohesion between the two parts than there is in the modern language.
Based on this type are two other formations, both dating to Early Modern Romanian as well. There is a perfect conditional formed with the conditional auxiliary, the infinitive of ‘be,’ fi, and the past participle, e.g., aş fi făcut ‘I would have done’ (lit., ‘I.would to.be done’), and there is also a formation with the auxiliary plus fi ‘be’ and a gerund, e.g., n-aţi fi având păcate ‘you would have no guilt’ (lit., ‘not you.would to.be having sins,’ Zafiu 2016: 25). This latter type is only “scarcely used in the contemporary language” (Zafiu 2013a: 42).
The etymology of the auxiliary is much debated, the competing hypotheses being that it involves reduced forms of ‘have’ or of ‘want.’Footnote 307 Zafiu 2016: 24 argues that additional dialectal evidence from Romanian of the Banat region and Istro-Romanian data, both involving conjugated forms clearly from vrea ‘want’ (e.g., I-R ręš face / Banat reaş face ‘I would do’), as well as “the numerous periphrases with vrea in O[ld][R[omanian,” settles the matter in favor of vrea ‘want’ as the source.Footnote 308
The corresponding analytic conditional forms in Aromanian and Meglenoromanian follow patterns seen in other languages, with an invariant originally 3sg form of the verb ‘want’ (present in Aromanian, imperfect in Meglenoromanian) plus a finite verb, the imperfect in the case of Aromanian. The dms is normally present in the Meglenoromanian formation, to judge from the forms given in Atanasov 1990: 226; 2002: 251, e.g., present conditional vrḙa si/sǎ fac ‘I would do,’ past conditional vrḙa si/sǎ am fat(ã). Capidan 1925b: 169 gives the conditional as ac(u) (< Mac ako) without the dms but tuku (< Mac) as requiring the dms (si). The presence of the dms also varies, in Aromanian, being given in Gołąb 1984a: 129–130 as a necessary part of the formation in Kruševo, e.g., va-s-videám ‘I would see,’ but being mostly absent from the examples in Bara et al. 2005: 195ff., e.g., v’ai biamu ‘I would drink’ but also v’ai si-askãpa ‘he would escape.’Footnote 309
As noted above, the past conditional offers a wide variety of periphrastic, analytic forms, with differences in different accounts that are not always easy to reconcile. Papahagi 1974: 66ff. gives other analytic types for the past conditional of Aromanian besides the type with va, namely invariant vrea (3sg imperfect of ‘want’) + imperfect indicative, e.g., vrea purtám ‘I/we would carry,’ invariant vrea with the dms s(i) and the imperfect indicative, e.g., vrea s’cădeám, and inflected imperfect indicative of avea ‘have’ preceded by the dms si and followed by the past participle, e.g., si-aveám cădzută, these last two meaning ‘I would have fallen; were I to have fallen.’ Vrabie 2000: 63–64 gives yet another periphrasis for the past conditional, namely invariant vrea with the present conditional (thus with the dms s(i)), e.g., vrea s-arcáre n-foc ‘he would have thrown (himself) into-the-fire’ (from aruc ‘throw’). Vrabie remarks (p. 64) that “this form of the conditional is more frequent in the Pindus area. In the North the usual past conditional consists of the auxiliary vrea + the imperfect,” with the dms, like one of Papahagi’s types, e.g., vrea s-cântái ‘you would have sung.’ He also notes that “there are also other types of past conditional, including ones in which the formant s- is absent,” and he gives as an example vrea lu da ‘(he) would have given,’ interestingly, with the present tense form da. These types with the dms are likely to be later developments out of the imperfect vrea + infinitive periphrasis seen in Early Modern Romanian, affected by the replacement of the infinitive (thus presumably a I.2 type, or variant thereof, from an earlier I.1 type, following the schema of §6.2.4.2).
All of this detail on the form of the conditional in Balkan Romance establishes clearly the Balkan character of the formation, with its innovative analytic character, the parallelism it shows with the future formation, the use of a verb of volition as an auxiliary, and the invariance of the auxiliary element across at least some of the relevant linguistic area, to mention just the key features. In fact, regarding the auxiliary, the observation of Zafiu 2013a: 42 is worth considering: “If the hypothesis of the auxiliary development [of aş] from the verb vrea ‘want’ is valid, then Romanian is the only Romance language [sic] whose conditional form developed from periphrases with a volitional verb.” Importantly, however, even if ‘have’ is the source of aş, the evidence of Aromanian and Meglenoromanian would single Balkan Romance out as special within the Romance family, given the shape and source of their conditionals. Moreover, as the material presented in §6.2.4.2 and the other related sections shows, these languages would not be the only Balkan ones with such a formation, so that – as with so much in the Balkans – the geography and the distribution of features across that geography are telling.
On the functional side of the ledger, Balkan Romance conditionals present a picture that similarly is not at all unusual from a Balkan point of view. We survey here the uses to which these conditional forms are put.
As Zafiu 2013a: 52 puts it, not surprisingly, the Romanian “conditional proper is characteristically used in conditional sentences, which are made up of a hypothesis plus the formulation of its possible consequences.” The same observation about the primary use of conditional forms holds for Aromanian and Meglenoromanian. As far as Romanian is concerned, the conditional form may appear in either the “if”-clause (the hypothesis, also known as the protasis) or in the “then”-clause (the consequence, also known as the apodosis), as the ensuing examples make clear. The basis for classification is the framework enunciated in §6.2.4.2.1; Romanian examples are from Zafiu 2013a and are marked as “Z” with a number indicating the example number in the source text, and Aromanian examples are from Vrabie 2000 and are marked as “V” with a page number (the orthography has been normalized in accordance with Cuvata 2006); the Meglenoromanian example in (6.107) is from Atanasov 2002, “A” with the page number indicated, and the one in (6.110) is from Atanasov 1990, “A’” with the page number indicated. See also Weigand 1896 and Bara et al. 2005 for examples. Special conditional forms of any sort, when they occur, are in bold face.
Present Expectative (Fulfillable)
Sentences that have a fulfillable apodosis and where the protasis sets forth conditions that are expected to hold occur in Romanian with the special conditional form, possible in both clauses, as (6.103ab) show. Aromanian examples are given in (6.104), where future tense is used in (a), given the immediacy of the prediction, and a conditional occurs in the protasis in (b).
a.
Dacǎ ai vrea, ai putea (Z, 59c) if cond.2sg want.inf cond.2sg can.inf ‘If you wanted, you could do it’ b.
Dacǎ ar vrea, ar pleca (Z, 64a) if cond.3sg want.inf cond.3sg leave.inf ‘If (s)he wanted (to), (s)he would leave’ c.
Dacă aş rămâne if cond.3sg remain.inf ‘If I stay … ’Footnote 310
a.
Ma va ti veádã tine cu míne, va ti vátãmã (V, 390) if fut you.acc sees you.acc with me fut you.acc kill.3sg ‘If he sees you with me, he will kill you’ b.
Si aspuneárii … minciúnã … âtsi dau viptul tut (V, 63) dms tell.cond.2sg lie you.dat give.1sg food all ‘If you were-to-tell a lie … I (will) give you all the food’
Hypothetical Fulfillable
Hypothetical fulfillable conditions have an apodosis that can be fulfilled and a protasis that sets forth hypothetical conditions. In such sentences, all the Balkan Romance languages can use their respective conditional forms, as (6.105) from Romanian, with a past conditional, (6.106) from Aromanian, and (6.107) from Meglenoromanian all show:
(6.105)
Dacǎ ar fi mers la mare, ne întâlneam (Z, 60) if cond.3sg/pl be.inf go.ptcp to sea us meet.impf.1pl ‘If (s)he/they had gone to the seaside, we would have met’
(6.106)
Cǎ se-aflãri vârã drac … sã s-toarnã nãpoi (V, 63) if dms-find.cond.3sg a devil dms him-turn.3sg back ‘if he came across (‘were-to-find’) a devil, he would turn back’
(6.107)
s-ou̯ vǎdḙám ódrina la vrḙámi, dms-it.acc wet.impf.1pl trellised_vine in time nu vrḙá s-ii̯ǎ úu̯a cǎta ampuvinítǎ (A, 278) neg cond dms-be.pl grapes so withered ‘if we irrigated the vine in time, the grapes would not be so withered’
Expectative Unfulfillable
Expectative unfulfillable conditionals have a past-looking protasis that puts forward an irreal (counterfactual) set of circumstances, and have an apodosis that is similarly rooted in an irreal world, thus making it unfulfillable. While the special conditional forms can occur in this type of conditional, as the examples below from Romanian (6.108), Aromanian (6.109), and Meglenoromanian (6.110) demonstrate, one interesting further possibility in Romanian, seen in (6.108a), is the occurrence of the simple imperfect tense in both clauses. This usage is thus consistent with the Greek evidence discussed in §6.2.4.2.4, where it is noted that the prevalence of imperfect tense forms in Greek conditionals puts the imperfect at the heart of conditional modality, more so than any other tense; this seems to be a distinctly Balkan property.
a.
Dacǎ voia, pleca (Z, 64c) if want.impf.3sg leave.impf.3sg ‘If (s)he had wanted (to), (s)he would have left’ b.
Dacǎ voia, ar fi plecat (Z, 64d) if want.impf.3sg cond.3sg be.inf leave.ptcp ‘If (s)he had wanted (to), (s)he would have left’ c.
Dacǎ ar fi vrut, pleca (Z, 64e) if cond.3sg be.inf want.ptcp leave.impf.3sg ‘If (s)he had wanted (to), (s)he would have left’
(6.109)
S-fúrim io tu lóclu a lui 99 di bǎrţate dms-be.cond.1sg I in place.the of his 99 of (measuring-)spans vrea s-lu hidzeárim tu loc (V, 63) cond.aux dms-him drive.cond.1sg into place ‘If I had been in his place, I would have driven him 99 feet down into the earth’
(6.110)
Acu ḱinise̯à mai̯ curún, vre̯a s-ii̯ǎ júnsu cásǎ (A’, 226) if move.impf.3sg more soon cond dms-be.3sg arrive.ptcp home ‘If he had left sooner, he would have arrived home’
Hypothetical Unfulfillable
This type seems not to have a formal realization in Balkan Romance. One further point about conditionals in Balkan Romance is that in Aromanian and Meglenoromanian, some of the conjunctions/complementizers that are used to introduce the protasis of conditional sentences are borrowings. For instance, ama in Aromanian ama că is most likely borrowed from Greek άμα ‘when, if,’ and Meglenoromanian acu is from Macedonian ako. See §4.3.3.4 and §6.2.4.3.1.2 on contact effects regarding such complementizers.
Finally, it must be noted that besides the uses seen in these various sorts of conditions, the Balkan Romance conditional serves other functions as well, which are treated in separate sections. See §6.2.4.2.8 on attenuation and optativity, and §6.2.5 on evidentiality, especially §§6.2.5.7–6.2.5.8.
6.2.4.2.3 Albanian
Unlike Balkan Slavic and Balkan Romance, but like Greek and Romani (leaving to one side borrowed particles such as Slavic bi), Albanian does not have a dedicated conditional per se, although, unique among the Balkan languages, it does have a dedicated optative. This optative, together with various modal uses of morphological, past tenses (all appropriately modified by items such as fut, dms, ‘if,’ etc.) constitutes the bulk of Albanian conditionals.Footnote 311 In general, Albanian follows the practice of using future imperfects and pluperfects (imperfect auxiliary, see Table 6.30) as conditionals, but owing to the fact that any past tense as well as any optative can be used in a conditional period – albeit some forms far more frequently than others – there is considerable room for pragmatic or stylistic variation (see examples (6.116) and (6.118abc) below). With Kramer’s 1986 framework as a starting point, the Tosk (and Standard) Albanian facts are surveyed here. Boretzky (2014) gives a thorough account of the Geg conditionals that use the past tense of ‘have’ + INF (= me + short ptcp). What is noteworthy in Boretzky’s account is that the current Geg constructions appear to be of relatively recent (early modern) origin. They are not well represented in Buzuku (1555, = Çabej 1968 in the list of references), and from this we can speculate that the current Geg situation is, in fact, also of Balkan origin, albeit via a different route, insofar as the imperfect of the future auxiliary, i.e., the imperfect of the ‘have’ future + dms equivalent (i.e., the Geg infinitive) is used for the conditional (cf. Boretzky 2014; Fiedler 2004). The Standard Albanian facts, which reflect Tosk, are given below.
Expectative Fulfillable
A typical Albanian expectative fulfillable has prs.sbjv in the protasis and fut in the apodosis, as in (6.111):
(6.111)
Po të shkosh ti, do të shkojë edhe ai. if dms go.sbjv.2sg you fut dms go.sbjv.3sg and he ‘If you go, he will go’
But the aorist can also occur in the protasis followed by a present in the apodosis:
(6.112)
Po s’ erdhe ti, as ne s’ vemi if neg come.aor.2sg you neither we neg go.prs.1pl ‘If you don’t come, we don’t go either’ (Newmark et al. 1982: 71)
It is also possible to have an optative in the protasis with a future in the apodosis:
(6.113)
Në paça kohë, do të shkoj if have.opt.1sg time fut dms go.prs.1sg ‘If I have time, I will go’ (Newmark et al. 1982: 108)
Expectative Unfulfillable
For expectative unfulfillables, a typical pattern is im-pluperfect subjunctive + future im-pluperfect:Footnote 312
(6.114)
Sikur të kishin luftuar mbretërit e Bizantit as.if dms had.impf.3pl fought.ptcp king.pl pc.nom.pl Byzantium.gen si Skënderbeu, Stambolli nuk do të kishte rënë as Skanderbeg.def Istanbul.def neg fut dms had. imp.3sg fall.ptcp në duar të turqve. in hand.pl pc.acc.indf.pl Turk.pl.gen ‘If the kings of Byzantium had fought like Skanderbeg, Istanbul would not have fallen into the hands of the Turks’ (Newmark et al. 1982: 318)
However, imperfect subjunctive + future im-/pluperfect is also possible:
(6.115)
Footnote 313Të mos të kisha në vatër time, dms mneg you.acc had.impf.1sg in hearth my.acc.f do të të kisha vrarë. fut dms you.acc had.impf.1sg kill.ptcp ‘If you were not living in my home, I would have killed you’ (Newmark et al. 1982: 75)
Hypothetical Fulfillable
Example (6.116) demonstrates the variety of pragmatic effects from different verbal bases in the protasis. This example has three hypothetical (potential) protases, with three different words for ‘if,’ three different tenses in the protasis, and a present-as-future in the apodosis. The first protasis has po + dms + prs.sbjv, the second has në + prs.opt, and the third has po + aorist. All three of the apodoses are, strictly speaking, fulfillable insofar as they could happen. In terms of pragmatics, however, the effect is one of dramatic build-up.Footnote 314
(6.116)
Atij që nuk të do, po t’ i japësh he.dat comp neg you.acc like.prs.3sg if dms him.dat give.sbjv.2sg majëzën e thoit, të rrëmben tip.acc.def pc.acc.def fingernail.gen.def you.dat grab.prs.3sg gishtin, në i dhënç gishtin, të merr finger.acc.def if him.dat give.opt.3sg finger.acc.def you.dat takes dorën, po i dhe dorën, të merr trupin … hand.acc.def if him.dat give.aor.2sg hand.acc you.dat takes body.acc.def ‘If you give the tip of your fingernail to him who does not like you, he will take your finger; if you give him the finger, he will take your hand; if you give him the hand, he will take your body … ’ (Newmark et al. 1982: 307)
(6.117)
nuk do të bënin gabim, po të kalonin neg fut dms make.impf.3pl mistake if dms pass.impf.3pl në ndonjë tavolinë tjetër. to some table.acc.indf other ‘They would not be making a mistake if they moved over to another table.’ (Newmark et al. 1982: 89)
Hypothetical Unfulfillable
The following three examples (6.118abc) illustrate the same hypothetical unfulfillable conditional period using three different constructions. The context is a brief article that appeared in an Albanian newspaper:Footnote 315 (6.118a) is the headline, (6.118b) is a report, and (6.118c) is a quotation. The headline, which is intended to grab the reader’s attention, uses po + dms + aorist in the protasis, and a colloquial future imperfect (do + imperfect without the dms) in the apodosis. The report, which is leading into the quotation, uses the colloquial future imperfect in both protasis and apodosis, thus engaging the reader’s attention for the actual quotation, which uses a standard Albanian imperfect subjunctive in the protasis and future imperfect in the apodosis.
a.
Bregu: PD do ishte ndryshe, po të qe Olldashi gjallë Bregu PD fut be.impf.3sg different if dms be.aor.3sg Olldashi alive ‘Bregu: the D[emocratic] P[arty] would be different if Olldashi were alive’ b.
Bregu tha se kjo forcë politike nuk do ishte Bregu say.aor.3sg comp this.f power political.f neg fut be.impf.3sg në këtë gjendje nëse Sokol Olldashi do ishte gjallë. in this.acc position if S.O. fut be.impf.3sg alive ‘Bregu said that this political force would not be in this situation if Sokol Olldashi were alive.’ c.
Po të ishte gjallë, Sokol Olldashi, PD-ja if dms be.impf.3sg alive S.O. PD-def.f.nom do të ishte ndryshe. fut dms be.impf.3sg different ‘If Sokol Olldashi were alive, the PD would be different.’
The perfect optative can also be used in the protasis as in example (6.119):Footnote 316
(6.119)
Në paçim punuar edhe ne kështu për ngritjen if have.opt.1pl work.ptcp and we thus for raising.acc.def e kooperativës, vaj halli. pc.acc.def cooperative.gen.def woe trouble.def.nom ‘If we too had worked like this in the setting up of the cooperative, forget it!’ (Newmark et al. 1982: 90).
Note also the Macedonian example cited in Footnote footnote 303 above, as it had this translation into Albanian: të mos isha president edhe unë do të shpërngulesha nga vendi ‘If I were not president, I, too, would emigrate from the country’ (lit., ‘dms mneg be.impf.1sg and I fut dms emigrate.impf.1sg from country.def.nom.M), showing dms + impf in the protasis and fut + dms + impf in the apodosis as in (6.118c).
6.2.4.2.4 Greek
As indicated in §6.2.4.2 and especially in Footnote footnote 290, the Ancient Greek way of marking conditional-like modality was very different from the Balkan type that has been described in preceding sections. Ancient Greek made considerable use of the modal/conditional particle ἄν and had a verbal mood, referred to as the optative, that fulfilled various modal functions pertaining to the oppositions relevant for a consideration of conditionals in later Greek, namely those identified by Kramer 1986: expectative/hypothetical in the protasis (if-clause), and fulfillable/unfulfillable in the apodosis (then-clause). In the discussion to follow, Kramer’s framework is taken as the point of reference.
Two clarificatory background points are in order here. First, as noted in §6.2.4.2, drawing on Horrocks 2010: 131, Postclassical Greek, starting in the Koine period and continuing into the modern era, makes use of the innovative periphrastic future tense formations with ἔχω ‘have’ and θέλω ‘want’ (on which see §6.2.4.1.1) as the basis for the formation of modal conditional forms that could express, for instance, hypothetical conditions; these periphrases replace earlier synthetic modals. Second, the Ancient Greek modal particle ἄν, which does so much work in that period in the expression of modality, while still quite present in the Greek of the New Testament, fades from use throughout the Postclassical period, perhaps helped towards desuetude by the breakdown of the old synthetic modality, and it survives in just a few uses in Modern Greek, if at all. The Modern Greek conditional conjunction αν ‘if’ contains Ancient Greek ἄν but only in a somewhat hidden way: αν ‘if’ is not from ancient modal ἄν directly but rather is from Ancient Greek ἐάν, a composite of εἰ ‘if’ with the modal ἄν, via a contraction of ἐάν to ᾱ́ν (note the long vowel), which then gives modern αν ‘if’ by regular loss of vowel length. Some modern uses of αν that resemble ancient modal ἄν could represent extensions of the ‘if’ meaning or could be legitimate survivals of the ancient form, as for instance in (6.120), with a modal sense of uncertainty:
(6.120)
Ό,τι κι αν λες, δε σε πιστεύω whatever and AN say.2sg neg you.acc believe.1sg ‘Whatever you might say, I do not believe you.’ (lit., ‘whatever and AN you-say, … ’)
The examples given here for Modern Greek with αν, and most others that one may encounter, represent the ᾱ́ν from ἐάν meaning ‘if.’Footnote 317 As in preceding sections, the different kinds of conditional modality patterns are presented, based on the framework of Kramer 1986; it can be seen that especially the imperfect tense, and the future marker θα and the pluperfect as well (see §6.2.3.3), do a considerable amount of conditional modal work in the modern language.
Expectative Fulfillable
In sentences with a fulfillable apodosis, where the protasis sets forth conditions that are expected to hold, Modern Greek can use a variety of tenses and moods in either clause. Examples (6.121abc) illustrate a few of the possible combinations, with perfective and imperfective presents and an overt future in the protasis and so also in the apodosis:
a.
Αν δουλεύεις σκληρά πάντα πετυχαίνεις if work.ipfv.prs.2sg hard always succeed.ipfv.prs.2sg ‘If you work hard you always succeed’ (Holton et al. 1997: 459) b.
Αν με φωνάξεις, θα έρθω if me.acc call.pfv.prs.2sg fut come.pfv.1sg ‘If you call me, I will come’ c.
Αν θα πάω στην Ελλάδα θα προσπαθήσω να if fut go.1sg to.the.acc.sg Greece.acc.sg fut try.1sg dms τον δω him.acc.sg see.1sg ‘If I (will) go to Greece, I will try to see him’ (Holton et al. 1997: 207)
Hypothetical Fulfillable
When the apodosis is fulfillable but the protasis sets forth conditions that are hypothetical, Modern Greek has a more restricted range of possible verb forms for both clauses. Either verb can be imperfect (i.e., imperfective past tense) and either can be pluperfect (i.e., past perfect, consisting of the past (imperfect) of ‘have’ and the perfect formative), although the verb in the apodosis is in either case preceded by the future/modal marker θα (thus giving a conditional or a conditional perfect). Examples (6.122abcd) from Holton et al. 1997: 208 are illustrative:
a.
Αν διάβαζες το γράμμα του θα καταλάβαινες if read.impf.2sg the.acc.sg letter.acc.sg his.gen fut understand.impf.2sg ‘If you were to read his letter you would understand’ b.
Αν είχες διαβάσει το γράμμα του if have.impf.2sg read.prf the.acc.sg letter.acc.sg his.gen θα καταλάβαινες fut understand.impf.2sg ‘If you had read his letter you would understand’ c.
Αν διάβαζες το γράμμα του θα if read.impf.2sg the.acc.sg letter.acc.sg his.gen fut είχες καταλάβει have.impf.2sg understand.prf ‘If you had read his letter you would have understood’ d.
Αν είχες διαβάσει το γράμμα του if have.impf.2sg read.prf the.acc.sg letter.acc.sg his.gen θα είχες καταλάβει fut have.impf.2sg understand.prf ‘If you had read his letter you would have understood’
Holton et al. 1997: 208 describe the differences in these sentences thus: “[These] sentences … imply that the subject of the ‘if’ clause has not read the letter and has not, therefore, understood.” They observe further, however, that (6.122a) “allows for the possibility that the wish of reading the letter and understanding may still be realized,” thus in the terms adopted here, an expectative fulfillable conditional. They make it clear, though, that in (6.122b–d) “where a pluperfect is used in one or both of the clauses, [the sentences] are more definitely counterfactual,” thus in the terms adopted here, a hypothetical fulfillable conditional, in that the outcome of understanding is achievable even under the hypothetical circumstances outlined in the apodosis.
Expectative Unfulfillable
An expectative unfulfillable conditional also utilizes a pluperfect with either a pluperfect modalized with θα (i.e., a conditional perfect) or a simple modalized imperfect (thus “conditional” in the more traditional sense of a blend of future and past reference), as in these synonymous examples, from Holton et al. 1997: 460:
a.
Aκόμη κι αν σου είχα πει την αλήθεια still and if you.gen have.impf.1sg say.prf the.acc.sg truth.acc.sg δε θα με πίστευες not fut me.acc believe.impf.2sg b.
Aκόμη κι αν σου είχα πει την αλήθεια still and if you.gen have.impf.1sg say.prf the.acc.sg truth.acc.sg δε θα με είχες πιστέψει not fut me.acc have.impf.2sg believe.prf ‘Even if I had told you the truth you wouldn’t have believed me’
The authors describe the situation related in (6.123) in these terms: “the meaning is that even if it had been the case that I had told the truth (which is not the case because I didn’t) it would still have been true that you would not have believed me.”
Hypothetical Unfulfillable
In the hypothetical unfulfillable conditionals, once again the imperfect occurs, combining with θα in the apodosis to give the conditional modality:
(6.124)
Aν το μωρό μπορούσε να μιλήσει, θα σου έλεγε … if the.nom.sg baby.nom.sg can.impf.3sg dms speak.3sg fut you.gen say.impf.3sg ‘If the baby could speak, it would say to you … ’
To sum up what emerges from these various conditional types, it should be noted that what the unfulfillable and the hypothetical cases have in common is that imperfect tense (διάβαζες, είχες, and καταλάβαινες in (6.122), είχα and πίστευες in (6.123a), and μπορούσε and έλεγε in (6.124)) figures prominently in each type, whether as the imperfect per se or as the imperfect of ‘have’ which combines with future/modal θα in the conditional perfect, and regardless of which clause is concerned. The imperfect tense thus lies at the heart of conditional modality, more so than any other tense, and as a comparison with the other languages surveyed here shows, this is a property that is distinctly Balkan in character.
Another feature of Greek conditionals that finds a parallel elsewhere in the Balkan languages is the fact that the subjunctive marker, the dms να by itself, can impart the conditional modality of a protasis without αν or εάν or any other overt word for ‘if,’ e.g., άμα.Footnote 318 That is, sentences such as (6.125) have fully conditional semantics (Holton et al. 1997: 458):
(6.125)
να τον δεις τώρα μετά την εγχείρηση του dms him.acc see.2sg now after the.acc.sg operation.acc.sg his.gen δε θα τον γνωρίσεις not fut him.acc know.2sg ‘If you were to see him now after his operation, you wouldn’t recognize him.’
This same usage is seen in Balkan Romani, where, as noted in §6.2.4.2.5, the native Romani conditional construction has only te, the Romani dms, in the protasis.
The expectative fulfillable examples also show two interesting morphosyntactic points. First, in (6.121b) above, the verb in the protasis (φωνάξεις) is in the perfective present, the category of verb that cannot stand on its own but must always occur with some governing element (see §6.2.2.3.3 and §7.7.2.1.3.2); the complementizer αν counts as one such governing element. Second, in the usual case with αν and the perfective present, there is no overt modal marker, just the bare verb with αν controlling it (though future θα is possible, as in (6.121c); see also Footnote footnote 244). Very occasional instances of the dms να can be found occurring in this context, as in (6.126), from a Greek website, but they sound ungrammatical to many speakers and some could even simply be errors:
(6.126)
είναι δυνατό κάθε παιδί να αγαπά να διαβάζει αν να του δείξει be.3sg possible each child dms love.3sg dms read.3sg AN dms him.gen show.3sg κανείς πόσο ενδιαφέρουσα είναι η διαδικασία someone.nom how.much interesting be.3sg the.nom.sg process.nom.sg του διαβάσματος the.gen.sg reading.gen.sg ‘It is possible for each child to love to read if someone shows him how interesting the process of reading is’Footnote 319
To turn now to a brief consideration of the history of this usage, here too the Balkan character of conditional modality in Greek becomes apparent. First, as noted at the beginning of this section, various periphrases with the past forms of ἔχω ‘have’ or μέλλω ‘be about to,’ and ultimately θέλω ‘want,’ somewhat parallel to the various future periphrases (see §6.2.4.1.1), occur from the Postclassical period into Medieval Greek. The considerable variation is surveyed in Holton et al. 2019: 1795–1814, where the observation, chronologically interesting from the Balkan perspective, is made that the ultimately predominating formation with ‘want,’ in particular the nonvolitional imperfect of ‘want’ + infinitive, was “well-established by the 13th/14th c. since [it is] even used in not-very-vernacular texts” (p. 1807).
As for the processes leading to these conditional forms, Horrocks 2010: 298–299 states that “bare subjunctives … [which] had begun as early as late antiquity to be strengthened by ἵνα [ˈina]/νά [na], … function[ed] not only modally but also as futures.” As a result, he says, “correspondingly, the ‘conditional’ (= ‘would (have)’ …) was expressed by a bare imperfect (i.e., the past of the present used as a future), and by the past-tense forms of the infinitival periphrases, i.e., εἶχα [ˈixa], ἔμελλα [ˈemela]/ἤμελλα [ˈimela], ἤθελα [ˈiθela], + infinitive.” The next step, leading to what he refers to (p. 299) as the “remodalization of the imperfect” tense, was for the “bare modal imperfect [to be] regularly strengthened, in a development modelled on the established use of νά [na] to mark a present as future/subjunctive in force.” The “strengthening” thus involved the use of the dms να. Horrocks 2010: 299 cites an early example of this sort, νά ἐδούλευα, from the twelfth-century Ptokhoprodromic poems:
(6.127)
ὡς σηκώτης να ἐδούλευα τὴν ἅπᾱσαν ἡμέρᾱν as porter .nom dms work.impf.1sg the.acc.sg whole.acc.sg day.acc.sg ‘I would have worked as a porter the whole day’ (Ptokhoprodromika III.182)
Horrocks (p. 300) sees this modality as connected with two other developments in the verbal system: the emergence in the late Byzantine period of θέλω ‘want’ as the primary auxiliary with an infinitive originally for the future tense (cf. §6.2.4.1.1) and the subsequent shift of ἔχω ‘have’ with the infinitive into the perfect system, starting with a pluperfect (see §6.2.3.3). The pluperfect was itself associated with modality, especially in conditional sentences, and he argues that due to the use of such modal expressions in the apodosis, να-clauses, especially with the imperfect, came to be used in the protasis, as a kind of balancing out of modality in the two connected clauses. The use of να for ‘if’ is found still today, as (6.125) above indicates.
Moreover, the primacy of θέλω-based futures had a consequence for modality with the imperfect, in two ways. First, just as a present of θέλω with an infinitive could signal futurity, the corresponding construction with past of θέλω could signal an anterior future, i.e., a conditional in the classic sense, and indeed such forms abound in Medieval Greek, e.g., ἤθελα γράψειν ‘I would write.’Footnote 320 Furthermore, just as various forms of θέλω with να plus a present verb could signal future, once the finite replacement for the infinitive affected the θέλω + infinitive future, the attachment of an imperfect of θέλω, i.e., 1sg ήθελα / 3sg ήθελε, to the regular infinitival replacement (να + a nonpast form) or to a modal να + imperfect was almost inevitable. Still, since the initial vowel of ήθελα/ήθελε is accented, there is no way by regular sound change to reduce ήθελα/ήθελε to a θ-initial form that could yield θα on its own. Thus, something more is needed, as noted in §6.2.4.2 and Figure 6.1, either “leakage,” i.e., a balancing, of tense across the forms to ensure an imperfect in the complement to ήθελα/ήθελε to which θα could ultimately be added, or an apparent flipping of past tense from the ‘want’ form to the complement, the “abrupt reanalysis” referred to in §6.2.4.2 (and see Pappas 2001), to give a presumed θέλει να έγραφα from which θα έγραφα could develop, following the same path from θέλει να to θα as in the future tense.
These developments were all taking place in a period and in a space in which there was contact among Greek speakers and speakers of other languages in the Balkans, at least Slavic, Romance, and Albanian. Given the similarity of the ways in which conditional modality came to be indicated in the languages of the region, it seems difficult to assume that these were entirely independent internal developments in each language; among other issues, one has to wonder whether the conditions for the reinforcement by a form of ‘want’ would have been present in each language. Moreover, given the synergy among the languages in other domains of grammar, it is equally hard to see how Greek alone, or any other language, for that matter, could be the source.
6.2.4.2.5 Romani
As indicated in §1.2.3.5, there are three major dialect groups of Romani spoken in the Balkans, and each of these has numerous individual subdialects. As Matras 2007 observes, Romani modal systems are particularly open to contact influences, and thus the expression of conditionals is quite varied. The account here, following §1.2.3.7, is representative of Skopje Arli (cf. Friedman 2017b), which shows significant Macedonian (and in some instances general Balkan) contact influences, but also has some similarities to Albanian (cf. also Friedman 2014c). The native Romani conditional construction uses only the dms in the protasis. Owing to the fact that the dms by itself can also have an optative-imperative force, Romani can use various borrowed words for ‘if’ – e.g., BSl ako, WRT eger – with or without the dms either to emphasize or to avoid ambiguity. Romani also has a native item: tejsi ‘if’ < te isi ‘dms is’ (cf. Albanian nëqoftëse ‘if’ < në qoftë se ‘if be.opt.3sg comp’). The dialects of Romani described here have borrowed Macedonian bi and have developed the typical Balkan conditional of the type fut + pst. The result is a quadripartite distinction that parallels that of Balkan Slavic, i.e., Expectative/Hypothetical and Fulfillable/Unfulfillable, as illustrated by examples (6.128–6.135). The fact that the Romani dialects considered here can use the dms with any tense and can also use the borrowed Balkan Slavic conditional marker bi means that there is more than one possibility for some types of conditionals.
Expectative Fulfillable
The ordinary construction is dms + prs, fut + prs:Footnote 321
(6.128)
Te mangen, ka khelen dms want.prs.3pl fut dance.prs.3sg ‘If they want, they will dance.’ (Jusuf 1974)
Like Albanian, Romani can also use the aorist after the dms in the protasis of a hypothetical fulfillable conditional sentence, with the apodosis in the present, as in (6.129):Footnote 322
(6.129)
Me te mangljan ka ingarav tut ki dučana leste. I dms want.aor.2sg fut lead.prs.1sg you.acc to.f shop he.loc ‘If you want, I’ll take you to his shop.’ (Jusuf 1974)
Hypothetical Fulfillable
For hypothetical conditionals, the borrowed particle bi, from Slavic, is used in the apodosis with the long form of the present. The protasis can be either the dms plus short present (6.130) or kana ‘when’ + bi + long present (6.131):
(6.130)
Te čingare man, me bi avava dms call.2sg.prs me.acc I cond come.1sg.prs ‘If you called/were to call me, I would come.’
(6.131)
Kana bi čingarea sine man, me bi avava when cond call.2sg.prs rem me.acc I cond come.1sg.prs ‘If you were to call me, I would come.’
Expectative Unfulfillable
For expectative unfulfillable conditionals, the protasis has the dms plus imperfect (6.132, 6.134) or pluperfect (6.133), i.e., the present or the aorist plus the remoteness marker, and the apodosis has fut plus impf, which may or may not include bi before the remoteness marker if the protasis has an imperfect:
(6.132)
Te ovel sine po džanlo ka džanel, sine po šukar. dms be.prs.3sg rem cmpv smart.m fut know.prs.3sg rem cmpv good ‘If he had been smarter, he would have known better.’
(6.133)
Te alo sine i daj, ka trajen sine ko adava moment. (RMS) dms come.pfv.3sg rem the.f mother fut quieten rem in that momentFootnote 323 ‘If their mother had come, they would have calmed down right away.’
(6.134)
Te ovel hine i daj lengiri, ka smirinol bi hine len. (RMS) dms be.3sg.prs impf the.f mother them.gen.f fut calm3sg cond impf them. ‘If their mother had been [there], she would have calmed him down.’
Hypothetical Unfulfillable
Like all hypothetical conditionals, the apodosis contains bi (cf. Boretzky 1993: 88, 91). In example (6.135), the protasis has the dms + short present (subjunctive) while in (6.136) the protasis has kana ‘when’ plus bi plus long present (indicative):
(6.135)
Te šaj sine o phuro te odgovorinel leske bi vakerel(a) dms can rem the.m old.m dms answer.3sg.prs him.dat cond speak.3sg.prs leske nešto him.dat something ‘If the old man could answer him, he would say something to him’ (i.e., but he’s asleep)
(6.136)
Kana bi čingarea sine man, me bi avava sine when cond call.2sg.prs rem me.acc I cond come.1sg.prs rem ‘If you were to have called me, I would have come.’
As can be seen from the foregoing examples, in Romani as in Macedonian (see §6.2.4.2.1 and (6.101)), hypothetical bi can also occur in expectatives.
6.2.4.2.6 West Rumelian Turkish and Gagauz
Taking Jusuf 1987: 94–102, Hafiz 1979, and Pokrovskaja 1964: 198–224 as starting points, it can be argued that in terms of the expression of conditionals, the differences between Balkan Turkish and Standard Turkish are, for the most part, phonological. Like Balkan Slavic and Balkan Romance, Turkish has a dedicated conditional morpheme (-sA, copular ise/-(y)sE), although unlike the Indo-European Balkan languages (except Judezmo), the Turkish noncopular morpheme attaches directly to the verb stem. The conditional copula is less likely to cliticize in WRT, consistent with the general tendency of copular forms that can be cliticized to remain independent in WRT.Footnote 324 The conditional copula can also modify or attach to the progressive (-yor), future (-AcAK), past (both mIş- and -DI), and the simple present (also known as the aorist, indefinite future, or gnomic present – Turkish geniş zaman ‘broad tense’ – in -r, negative -z). In addition to these constructions, ise can encliticize to past copulas idi and imiş, which can then be added to all of the aforementioned forms, except that imiş + ise cannot follow the DI-past. Furthermore, the clitics idi and imiş can attach to the conditional base and also have conditional effects when cliticizing to the future or simple present. The complexities of this system are beyond consideration here (see Lewis 1967: 96–136), but the main points are these:
(i) By the time Turkish arrived in the Balkans, the modal system as we know it was already more or less in place (cf. Erdal 2004: 270–271, 521). However, some specific changes, such as the addition of the particle çak to the base in -(y)A to form the future tense owing to the polysemy of -(y)A, took place some time during the history of Ottoman Turkish (Erdal, p.c.).
(ii) The addition of the DI-past copular (idi) to the simple present and the future for apodoses of hypothetical and expectative unfulfillable conditions is reminiscent of the use of imperfects (or dedicated conditionals in BSl and BRo) and anterior futures to form such conditionals in various Indo-European Balkan languages.
(6.137)
Yağmur yağ-ma-sa-ydı gez-me-ye gid-er-di-k rain rain-neg-cond-pcop walk-inf-dat go-prs-pcop-1pl ‘If it weren’t raining, we would go for a walk’
(6.138)
Yağmur yağ-ma-sa-ydı gez-me-ye gid-ecek-ti-k rain rain-neg-cond-pcop walk-inf-dat go-fut-pcop-1pl ‘If it hadn’t rained, we would have gone for a walk’
On the one hand, when looked at in isolation, the parallels to other Balkan languages are striking, and the relevant developments in those other languages took place, for the most part, during the period of contact with Turkish. On the other hand, when the Turkic systems, including the WRT and Gagauz systems, are examined in their full context, such similarities as there are appear to be typological. In the end, a cautious assessment would be that those parallels that do exist between the Indo-European and Turkic conditional expressions could at least have been mutually reinforcing, a point of relevance for the consideration of Judezmo in §6.2.4.2.7.
6.2.4.2.7 Judezmo
Judezmo has a synthetic future, a synthetic conditional, and also conditional uses of the synthetic imperfect subjunctive. Thus expectative fulfillable and hypothetical fulfillable conditionals are, or at least can be, the same as in Spanish. In Judezmo, however, such usages are bookish (Montoliu & van der Auwera 2004: 463). Closer to the Balkan type of conditional is the use of the imperfect indicative in both the protasis and apodosis of a hypothetical conditionals, which is possible in both languages, but is considered sub-standard in Spanish and normal in Judezmo (Montoliu & van der Auwera 2004: 462–464). As with other Balkan languages, Judezmo has a number of possibilities for expressing the same type of conditional, as indicated by the variant expressions in examples (6.139a–c), all of them hypothetical (from Montoliu & van der Auwera 2004: 463–464):Footnote 325
a.
Si me yamaran iriya if me.obl call.impf.sbjv.3pl go.prs.cond.1sg b.
Si me yamavan ya iva if me.obl call.impf.ind.3pl ptcl go.impf.ind.1sg c.
Si me yamavan iriya if me.obl call.impf.ind.3pl go.prs.cond.1sg d.
Si me yamavan ya iva a ir if me.obl call.impf.ind.3pl ptcl go.impf.ind.1sg to go.inf ‘If they called me/were to call me, I would go.’
Example (6.139a) is bookish, examples (6.139bd) use a particle whose lexical meaning is ‘already’ but which tends to simply mark apodoses in conditional periods, and example (6.139c) offers yet another possibility. An unambiguous expectative unfulfillable (past irrealis) can be formed with a pluperfect and imperfect where the protasis is pluperfect, while the apodosis is still imperfect (Montoliu & van der Auwera 2004: 465):
(6.140)
Si me aviyan yamado ya iva if me.obl have.impf.ind.3pl call. ptcp ptcl go.impf.ind.1sg ‘If they had called me, I would have gone’
In terms of Balkan contact phenomena, Judezmo does not seem to display the intersection of morphological futurity and anteriority that is the hallmark of Balkan conditional convergence, although the change of ya ‘already’ into a kind of conditional marker is certainly a step in that direction. Judezmo also displays a tendency to use indicative forms rather than synthetic modals in modal contexts where both Modern Spanish and Old Spanish do not use indicatives (Montoliu & van der Auwera 2004: 467). Such changes took place in a Balkan context and are consistent with Balkan usage. Montoliu & van der Auwera 2004: 472 provide a table of nine functional and structural parallels in conditional formation and interpretation comparing Judezmo with Greek, Turkish, and Modern Spanish, among which features various uses of indicative pasts figure prominently. Judezmo agrees with Greek and/or Turkish and not with Modern Spanish in eight out of the nine features. The ninth and only feature where Judezmo agrees with Modern Spanish and not Greek and Turkish is the use of subjunctive imperfects in the protasis and conditional presents in the apodosis. Overall, then, in terms of conditional constructions, it is fair to say that the situation in Judezmo is consistent with its status as a Balkan language: there are clear parallels with other Balkan languages that have arisen precisely in the context of and during the time of contact with the other Balkan languages, and these features are not the same as those found in today’s Spanish.
6.2.4.2.8 Attenuated Conditionals and Modal Aorists
Attenuated conditionals are sometimes similar to impersonal optatives (§6.2.4.3.2.1) insofar as they involve an independent use of a clause introduced by an element that is normally used for subordination, a phenomenon that has been called “insubordination” (Evans 2007; Evans & Watanabe 2016), and sometimes expresses a wish. In general, attenuated conditionals can be understood as protases that assume but do not enunciate an apodosis, as in the following two Greek examples where (6.141) expresses a wish as an unfulfillable hypothetical protasis (basically, ‘if X had happened, X would have been a good thing’) and (6.142) expresses a question as a fulfillable hypothetical protasis (if X happens/were to happen, what would Y do?):
(6.141)
αν έβλεπες την Ελένη με το φόρεμα του if see.impf.2sg the Helen.acc with the.acc.sg dress.acc the.gen.sg χορού! dance.gen.sg ‘If you could only have seen Helen in her ball dress!’
(6.142)
Κι αν ξαφνικά μας δούνε? and if suddenly us.acc see.pfv.prs.3pl ‘What if they see us suddenly?’
In (6.141), να could replace αν, and either clause could be preceded by κι ‘and.’Footnote 326
Similarly, Romanian can use either de ‘if’ + conditional or imperfect or dacă ‘if’ + imperfect or past conditional in hypothetical unfulfillable independent protases as in (6.143a–d). These usages are similar to a plain conditional or an unsubordinated subjunctive in wishes and curses (6.143e):Footnote 327
a.
De-aş dormi puțin if-have.cond.1sg sleep.inf little ‘If only I’d slept a little!’ (Zafiu 2013a: 50) b.
De rămâneai … ! if stay.impf.2sg ‘If [only] you had stayed … !’ c.
Dacă ştiam if know.impf.1sg ‘If [only] I had known!’ (Cojocaru 2003: 143) d.
Dacă aș fi știut dinainte (= Să fi stiut dinainte) if cond be known.ptcp before dms be known.ptcp before ‘If only I had known before! (= I wish I had known before!)’ e.
Bătu-l-ar Dumnezeu (= bată-l Dumnezeu) punish.inf-him-cond God.def punish.sbjv-him God.def ‘May God punish him᾽ (Vasilescu 2013: 392) (= subjunctive without dms, p. 387)
In Balkan Slavic the dms + old perfect or imperfect can be used, although the old perfect is preferred for the first person owing to the nonvolitional implications, as in (6.144a) from Macedonian, while in Romanian a past subjunctive can be used, as in (6.144b):
a.
Da sum znael! (= Da znaev!) (Mac) dms am.1sg know.lf dms know.impf.1sg b.
Să fi știut (Rmn) dms be know ‘If [only] I had known!’
Macedonian also uses kamo ± da to express an unfulfillable wish, so that the dms plus verb can be omitted, e.g.:
(6.145)
Kamo (da e) srekja whither dms is lucky ‘One should be so lucky!’Footnote 328
Bulgarian here has dano, which expresses a hope rather than an unfulfillable wish, e.g.:
a.
Dano uspeeš DANO succeed.2sg ‘May you/if only you will succeed’ b.
Šte pătuvam do tam, dano go sreštna fut travel.1sg to there DANO him.acc meet.1sg ‘I will travel there, may I meet him!’
Here we should also note a usage in Bulgarian (and marginally in Macedonian) of the type da ne bi (lit., ‘dms + neg + be.3sg.ipfv.aor’) with the meaning ‘lest’ or ‘it shouldn’t happen.’ The collocation da ne can also occur in Bulgarian with an aorist in an apprehensive meaning, da ne nastina ‘lest you catch cold/don’t catch (a) cold’ (lit., ‘dms + neg + catch.cold.2sg.aor’) (Mitkovska et al. 2017).
Also in Balkan Slavic (and Aromanian – using keški/keške [Aro cheshchi, cheshche] – but not elsewhere), unfulfillable independent protases of the type ‘If only … ’ can use keški ‘if only’ from Turkish keşke ‘idem.’ This particle is glossed as Blg po-dobre bi bilo da / Mac bi bilo podobro da … ‘it would be better if … ’ as in the following Macedonian examples:
a.
Keški da ostanev i jas tamu! [nemaše da se sluči taa nesrekja] if only dms stay.impf.1sg and I there [neg.fut.pst dms intr happen.3sg.prs that accident] ‘If only I had stayed there! [that accident would not have happened]’ b.
Keški da e taka if.only dms is.3sg so ‘If only it were so!’
These sentences can be compared with Turkish keşke with a conditional + DI-past as in:
(6.148)
Keşke öl-me-sе-ydi-n! if.only die-neg-cond-pfv-2sg ‘If only you had not died!’
In Turkish, as in Balkan Slavic, the conditional without the particle can be used in such expressions.
As seen in §§6.2.4.2.3 and 6.2.4.2.5, the aorist can be subordinated to the dms in Albanian and Romani in order to render conditional protases. See also example (6.232) for Aromanian; §6.2.5.8 treats the Aromanian subordination of the aorist to both dms and fut. Both Greek and Macedonian (but not Bulgarian) have specific contexts that permit dms + aorist and Greek also has fut (θα) + aor. As Asenova 2002: 235 states for Greek – and as is also true for Macedonian – these are not conditionals per se.
The Greek constructions are attenuated modals that express presumption, supposition, or dubitative surprise, and they correspond to fut or dms plus the l-perfect in Balkan Slavic, as in (6.149) and (6.150).
a.
Θα το άκουσες (Grk, Householder et al. 1964: 106; Holton et al. 1997: 229) fut it.acc hear.2sg.aor b.
Šte si go čul (Blg, Asenova 2002: 235) fut be.prs.aux.2sg it heard.lf.m ‘You must have heard it [somewhere]’
a.
Λες εγώ να τα έσπασα τα πιάτα (Grk) say2sg I.nom dms them.acc break.aor.1sg the.pl.acc dish.pl.acc b.
Jas da sum gi skršil sadovite (Mac) I.nom dms be.prs.aux.1sg them. acc break.lf.m dishes.def.pl ‘Me, could I have broken the dishes?!’
The Greek presumptive/suppositional usage of fut + aor and the dubitative/surprise use of dms + aor, corresponding to uses of the inherited Balkan Slavic perfect, bear some similarities to the Romanian use of fut, dms, and cond plus fi ‘be’ plus gerund or past participle in forming the presumptive mood. The Romanian usage, however, despite its complexity of meaning and polysemy (especially with the past participle), is closer to the evidential types of usage found in Balkan Slavic (which involve the inherited perfect), Albanian, Turkish, and to a more limited extent Aromanian, Meglenoromanian, and Romani, and so it is treated in §6.2.5. To this we can add Adamou’s 2012b observation that in the Macedonian dialect of Ajvatovo (Grk Liti) a suppositional use of the type fut + aor in the meaning ‘must’ve’ has been calqued from Greek into the local Macedonian dialect. In Standard Macedonian and the dialects on which it is based, similar suppositions are rendered by fut + dms + impf, e.g., kje da beše početokot na školska godina ‘it must’ve been the beginning of the school year’ (lit., ‘fut dms be.3sg.impf beginning.def of school year’).
There is another use of dms + aor that has affective nuances comparable to some extent to the Greek use of dms + aor, namely the use in Macedonian of kako ‘like, as, how’ + dms + aor to mean ‘as if’ but with a specific nuance of ‘apparently.’ In this it is like an admirative, which, as seen in §6.2.5 is, underlyingly, an infelicitous dubitative. That is to say, surprise of the type ‘well, apparently, although I would not have believed it, it turns out that … ’ is the inverse of ‘well, I do not believe that … ’ In the Greek of (6.150) and the Balkan Slavic equivalent, the breaking of the dishes might or might not have happened, but the speaker finds it unbelievable regardless. Macedonian kako + dms + aor is illustrated by (6.151ab). While Balkan Romance does not permit the aorist after the dms, it does after cum de ‘as if’ where either an aorist or a perfect can occur, with an exclamatory admirative-dubitative meaning, as in (6.151c):Footnote 329
a.
Betonov kako da pukna, a?! concrete.def like dms crack.aor.3sg q ‘Well, [it appears] as if this concrete has cracked!’ b.
Filmskiot festival “Brakja Manaki” kako da stana film.def festival Brothers M. like dms become.aor.3sg omileno mesto za skandali favorite place for scandals ‘It appears the film festival “Manaki Brothers” has become a favorite place for scandals.’ c.
Cum de făcui / am facut greșeală așa as if do.1sg.aor / do.1sg prf mistake thus ‘How can I have made such a mistake?’
Part of the attenuation is in the lexical meaning of kako da ‘as if,’ but the co-occurrence with a synthetic aorist is specifically Macedonian in Balkan Slavic. (Greek can have σαν να ‘like dms’ + aor, but Greek permits dms + aor elsewhere, too; cf. Footnote footnote 189, Chapter 7, and Bužarovska 2006.)
For kako da stana ‘as if it has become/happened,’ Bulgarian uses the indicative complementizer če with the old perfect, e.g., kato če e stanal ‘as if it had happened’ or with a question word + aorist, e.g., kato če li stana ‘how [on earth] did it happen?’ or a lexical verb izgležda stana ‘it-appears it-became/has-become.’Footnote 330 As indicated above, Balkan Romance also has an equivalent construction but does not permit the dms with the aorist.
Albanian has the option of using an aorist-pluperfect in similar constructions:
(6.152)
Si të qe aty mbërthyer … / Si me gozhda të qe ngulur … as dms be.aor.3sg there fixed.ptcp as with nails dms be.aor.3sg thrust ‘As if you were fixed in place there … / As if you had been nailed down … ’ (Gjoka 2011: 131)
These are fixed usages rather than fixed phrases.
We can also note the use of fut + aor in Albanian to express a kind of irreal conditional wish, as in (6.153), from a translation of Shakespeare’s Othello (Act 3, scene 3):
(6.153)
Më mirë do të desha t’ isha një bretkosë (Noli 1916: 94) more good fut dms want.aor dms be.1sg.impf a toad ‘I had rather be a toad …’
Here the Albanian more literally would be ‘I would rather that I were a toad …,’ where the fut + aor of ‘want’ expresses ‘would rather.’ See also §6.2.5.5 on the use of imperfect and perfect admiratives with the dms, which are basically past subjunctive irreal conditionals.
Finally, we can note that BCMS permits dms + aorist for irreal conditionals. While the construction is now unusual or archaic, native speakers still accept it. This is a modal aorist usage similar to that in Albanian illustrated in example (6.118). Gołąb 1984b, quoting from Stevanović 1969: 857, gives an example based on Nazor 1908: 3, which we cite here in its correct entirety: Da ga ne udariste, bio bi lupnuo s vama o pećinu, bacio vas u oblake ‘If you hadn’t hit him, he would have crashed into the rocks with you and thrown you up to the clouds’ (lit., ‘dms him.acc neg hit.2pl.aor, be.m.lf cond throw.m.lf with you.pl.ins on rock.acc, throw.m.lf you.pl.acc in clouds.acc’). While normal BCMS usage today would require the unmarked past (old perfect) in the protasis, and some speakers feel that the negation makes the sentence more acceptable, still it must be noted here that dms + aorist as an ordinary irreal conditional is limited to BCMS and Albanian.
6.2.4.2.9 Conditionals: Summary
Conditionals in many respects are the stepchild of Balkan linguistics. On the one hand, they provide a ready pendant to their “big sibling,” the future tense, in that the two categories show many of the same parameters of variation – selection of ‘want’ versus ‘have’ as a governing auxiliary, invariant particle versus inflected auxiliary, presence or absence of modal subordinating marker (dms), use of an infinitival as opposed to a finite main verb, and so on – as well as semantic affinities regarding modality and irrealis states. Moreover, in many instances there is overt, and striking, parallelism between the two formations. On the other hand, while the future tense has attracted considerable scholarly attention for almost 200 years, dating back to Kopitar 1829, conditionals have generally received less attention, and there is still more to be studied about them.
Nonetheless, the discussion in the preceding sections, treating the individual languages as to the form and function of their conditionals, makes it clear that there is a high degree both of complexity in usage and of variability in formation with conditionals, in ways that can lead it to diverge from the future. Thus, while some theoreticians might argue that the future is not a tense per se but rather is modal in nature, given the uncertainty of actually “knowing” what the future might bring, the same reasoning cannot apply to classic irreal conditionals, as they necessarily blend both a modality of uncertainty with overt time reference, i.e., past time, and are thus “tensed,” as the classic characterization of the conditional as a “future in the past” indicates.
With regard to the specifically Balkan side of the conditional, there definitely does seem to be a Balkan character to some aspects of the conditional across the various languages. This allows for the conclusion that – just as with the future – it is fair to talk about conditionals as representing a genuinely convergent feature, i.e., a true Balkanism. And, there is a key geographic dimension to support this as well.
There is, for example, the recurring appearance of the imperfect tense on the main verb in various conditional formations and uses, coupled with the striking geographically compact clustering of Albanian, Greek, Macedonian, and Balkan Romance, in particular Aromanian – as well as co-territorial Romani dialects – as languages showing that feature. The imperfect thus occupies a special place in the realization of Balkan conditional modality and demonstrates that in these languages it serves to carry the modal value of the conditional, while at the same time, rooting it clearly, from a morphological standpoint, in the past tense.
Moreover, it is precisely in this cluster of languages, representing the Central Balkans, that an interesting development in the form of the conditional can be found. As noted especially in §6.2.4.2.4 about Greek, following Pappas 2001, but relevant for other languages as well, there was a shift between the pieces of the periphrastic conditional with regard to where the past tense marking resides morphologically. For instance, a past tense of ‘want’ with an (untensed) infinitive served as a conditional in Medieval Greek, e.g., ἤθελα γράφει ‘I would write’ (ἤθελα as past tense of θέλω, γράφει as the infinitive), yet the Modern Greek conditional is θα έγραφα, with an (untensed) particle (most likely based on the present tense of θέλω, however) and the imperfect tense form of the main verb. There was thus a “flipping” of the tense marking from the auxiliary to the main verb at some point. As pointed out in §6.2.4.2 (and see Figure 6.1), this flipping must be characterized as a rather abrupt “leap,” since it is not a straightforward “micro-step” – reanalyses generally are abrupt, to be sure, and involve a shift from an existing status quo to some innovative new structure and/or form, but this change involves far more than a minor adjustment, a change in one feature or the like; rather, it involves an uprooting and reassignment of features across the two parts of the periphrasis, as the dotted lines in Figure 6.2 are meant to suggest.
Reanalysis in Greek conditional
This shift is somewhat unusual in its magnitude (hence the characterization of it as a “leap”), and yet it occurred. Importantly, though, the same flip can be identified in the development of the conditional in Tosk Albanian, Aromanian, Macedonian, and Romani, where the conditional is formed with the future particle (do, va, kje, ka respectively) with the imperfect past tense.Footnote 331 The reanalysis clearly happened once, perhaps in Greek, though it is not important which language innovated first. However, since Macedonian has a more gradual transfer attested dialectologically (see Footnote footnote 331), Balkan Slavic, or at least Macedonian, may have followed a different path. Given that this tense-flipping is somewhat unusual, is it reasonable to assume it happened independently in four different languages? Occam’s Razor – entia non sunt multiplicanda praeter necessitatem ‘Entities are not to be multiplied beyond necessity’ – would have it here that the flipping should be posited as occurring just once, and given the known fact of significant contact among speakers of these central Balkan languages, evident, for example, in the positioning of clitics (see §5.5.3 and §7.3.3) and the template for the verbal complex (see §7.4.1.2.3), contact can be held to be responsible for the appearance of the same reanalysis in this geographic cluster of languages. On the other hand, it may be that the gradual development of Balkan Slavic influenced the Greek “flip.” For both Albanian and Balkan Romance we lack the medieval and, apparently, dialectological resources to be sure.
Regardless of whether we have a “leap” analysis, as in Greek, or a gradual one, as in Macedonian, then, the appearance of the particular form of the conditional in Tosk Albanian, Aromanian, Greek, and Macedonian, and Romani can be taken as a core Balkanism. At the same time, the Bulgarian, and, to a lesser extent, BCMS developments, as well as those in Romanian, Meglenoromanian, and Geg Albanian, represent convergent innovations vis-à-vis the respective inherited situations (to the extent that these can be reconstructed), but innovations that represent a structural remove from the core Balkanism, although see Boretzky 2014 on the Balkan dimension of the Geg parallels.
6.2.4.3 Volitionals
Following Ammann & van der Auwera 2004, the term “volitional” is used here for expressions of desire, which include wishes, appeals, commands, prohibitions, blessings, curses, and more. Ammann & van der Auwera 2004: 294–297 distinguish five types of volitionals based on occurrence with the category of person: (1) optative, for all three persons, (2) imperative, 2nd person only, (3) hortative, 1st and 3rd persons only, (4) exhortative, 3rd persons only, and (5) cohortative, 1pl only.Footnote 332 For the purposes here, the discussion is organized as follows: synthetic optatives (§6.2.4.3.1), analytic optatives (§6.2.4.3.2), analytic hortatives (§6.2.4.3.3), and imperatives and prohibitives (§6.2.4.3.4). Moreover, we can add a sixth category, namely impersonal optatives (§6.2.4.3.5). These are third person expressions of desire that are not directed at a specific third person. To some extent, these optatives overlap with attenuated conditionals discussed in §6.2.4.2.8. The main difference is that attenuated conditionals, like optatives, can be in any person, whereas impersonal optatives, like exhortatives, are limited to the third person.
6.2.4.3.1 Synthetic Optatives
Synthetic optatives in the Balkans are limited to Albanian, Turkish, and some of those Romani dialects that have Turkish conjugation (see §6.2.1.1.1 above and Friedman 2013b). The overwhelming majority of Albanian optatives (90 percent in Alekseeva 2012) occur in blessings, curses, and related expressions of desire such as ju bëftë mirë ‘bon appétit’ (lit., ‘to.you may.it.do good’ = Mac na zdravje ‘to [your] health,’ also said when encountering people who are eating or from a server at the end of a meal), or the polite reply shëndet paç ‘may you have health’ (lit., ‘health may.you.have,’ corresponding exactly to Macedonian zdravje da imaš). About 10 percent of the optatives in Alekseeva’s corpus function as hypothetical conditionals in the protases after [edhe] në (‘even if’). In principle, the synthetic optative can always be replaced by an analytic optative, but in contexts where a synthetic optative is expected, such usage is unidiomatic. The Turkish optative (also called subjunctive) has forms for all three persons, unlike the imperative, which is only 2nd/3rd person (Lewis 1967: 132–137; Göksel & Kerslake 2005: 83, 91). In practice, however, the third person optative has been replaced by the third person imperative, except in a few fixed expressions and the formulation of cautious questions, e.g., evde mi ola ‘might he be at home?’ (lit., ‘home.loc q be.opt,’ Lewis 1967: 132). Functionally, the 1st person optative and 3rd person imperative form a hortative. Optative 2nd persons are used for reported requests or commands (Lewis 1967: 133). Both Albanian and Turkish have past optatives (the Albanian using an optative auxiliary + participle, i.e., an optative perfect, the Turkish using -DI or -mIş) for wishes or conditionals that are unfulfillable (or, rarely, simply past). Some Romani dialects with Turkish conjugation, most or all in eastern Bulgaria, can use the innovating Turkish optative (i.e., 3rd person from the imperative) with or without the Romani dms (te). (See §7.6.2 on negated optatives in Albanian.)
6.2.4.3.1.1 Perfects as Optatives
In Balkan Slavic, as well as the rest of South and West Slavic, the old perfect can function as an optative, e.g., BSl dal ti Bog dobro ‘may God grant you well/good,’ BCMS živ[j]eli! ‘to your health!’ (lit., ‘may [we/you] live’). Vaillant 1966: 97 attributes such uses of the l-participle to an elliptical optative composed of da plus the conditional (3sg bi plus l-participle).Footnote 333 This usage shows up in the use of bilo to form indefinite pronouns, as well as in the conjunction bilo … bilo … ‘whether X or Y,’ which corresponds exactly to standard Albanian qoftë … qoftë … (optative of ‘be’; cf. English be he alive or be he dead; Friedman 2012b). See also §6.2.4.3.2 on dms + perfect as optative.
6.2.4.3.1.2 Balkan Romance Subjunctive in an Optative ExpressionFootnote 334
Relevant to this discussion too is the way in which certain optative forms in Albanian seem to have played a role in an innovation within the Balkan Romance verbal system. In particular, in Romanian, the Latin perfect subjunctive, e.g., cantaverim ‘I might have sung,’ has been replaced functionally by a periphrastic form called (e.g., in Cojocaru 2003: §4.2.4) the “optative-conditional,” especially in the past tense. This form consists of the special conditional auxiliary form – 1sg aş, 2sg ai, 3sg ar, etc. (see §6.2.4.2.2) – with the infinitive, and in the past by aş together with the invariant modal auxiliary fi followed by the infinitive, e.g., aş cânta ‘I would sing (if given the chance),’ aş fi cânta ‘I would have sung.’ This most likely is an innovation that post-dates Common Balkan Romance, as it is not found in Aromanian.
In Aromanian, however, a form fure occurs which continues the Latin perfect subjunctive of ‘be,’ fuerit (3sg), ‘it might have been.’ It has an innovative function compared to its Latin source, and that is where the Albanian optative comes in.
Aromanian fure occurs in the form fure-că,Footnote 335 which when used by itself means ‘if,’ and when occurring paratactically conjoined with itself gives the alternative disjunctive meaning ‘whether … or,’ as in fure-că-i bărbat i fure-că-i mul’are ‘whether it be a man or a woman’ (Capidan 1932: 511). This latter use is innovative compared to Latin, where sive … sive (rarely seu … seu), of demonstrative origin, is used. Moreover, it appears to be innovative within Balkan Romance, since Romanian uses the subjunctive of ‘be’ in this construction, fie … fie (cf. French soit … soit), and shows something very different for ‘if,’ dacă.
Interestingly, fure-că … fure-că corresponds to the standard Albanian qoftë … qoftë, which consists of the optative of ‘be’ doubled. The single use of fure-că for ‘if’ is matched by single Albanian qoftë in the composite conjunction në qoftë se, where se corresponds precisely to Aromanian că (see also §4.3.3.4).Footnote 336 An apparently related development is seen in Balkan Slavic, where both Macedonian and Bulgarian and, to a lesser extent BCMS, show the neuter l-participial form of ‘be,’ doubled, i.e., bilo … bilo, in the meaning ‘whether … or’; this is due to the innovative replacement of the 3sg imperative by the verbal form in -l with an optative meaning, a usage which underlies the BCMS toast živ[j]eli ‘let.us/may.we live’ mentioned above; however, unlike Albanian and Aromanian, bilo does not occur singly in an expression for ‘if.’
The Albanian optative is based on a past-tense (aorist) stem, and seems to date, partly at least, from post-Roman times, inasmuch as the -f-, according to Sh. Demiraj 1985: 896, 1129, may have originated from the hiatus consonant seen in vocalic stem aorists (e.g., puno-v-a ‘I worked’).
These facts mean that both Common Albanian and the Latin dialects that gave rise to Aromanian developed similar modally based expressions, and used them in similar conjunctional constructions. Moreover, these developments were happening at a time in the Balkans when speakers of Romance and of Albanian, as well as Slavic speakers, were in contact with one another and were all significantly restructuring their verbal systems. A circumstantial case can be made, therefore, for contact effects underlying the innovative developments with qoftë, fure, and bilo.
What makes such an account all the more compelling is that there is evidence for the calquing of a conjunction for whether … or. In particular, in the Albanian dialects of North Macedonia, the optative-based qoftë … qoftë has been replaced by the present admirative qenka … qenka, based on the model of Macedonian bilo … bilo. What has made bilo … bilo a suitable model is the fact that even though it matches qoftë … qoftë, bilo also has a use as a nonconfirmative (see §6.2.5), in which case it matches the Albanian admirative qenka. This latter matching, extended to the former, would allow for the innovative substitution of qenka for qoftë, thus giving qenka … qenka. This substitution was surely aided by the fact that the optative in Albanian, though productive and still clearly a living category within the verbal system, is nonetheless limited to mostly to fixed phrases, e.g., rrofsh! ‘thank you’ (lit., ‘may you live’) or me nder qofsh ‘you’re welcome’ (lit., ‘with honor may.you.be’) and various formulae, blessings, and curses (including obscenities) that are very much tied to the desiderative meaning of the optative. The conjunction use of qoftë thus falls outside of the more common scope of the optative, and so would be ripe for being replaced.
6.2.4.3.2 Analytic Optatives
The analytic optative is formed in all the Balkan languages by means of the dms, normally with the present tense (or subjunctive for Romani, Balkan Romance, and Judezmo), although other tenses can also occur, as described below. Unlike independent subjunctives, e.g., in French (Honi soit qui mal y pense ‘May he be ashamed who thinks evil of it’), the “insubordinate,” i.e., independent, use of the dms is a basic structural feature of the Balkan languages (Friedman 1985a; Ammann & van der Auwera 2004). Table 6.27 sums up the Balkan situation.
| Romani | te | hramonavFootnote 337 |
| Albanian | të | shkruaj |
| Greek | να | γράφω |
| Bulgarian | da | piša |
| Macedonian | da | pišuvam |
| Torlak BCMS | da | pišem |
| Romanian | să | scriu |
| SDBR | s(i)~ś | scriu |
| Judezmo | ke | eskrivа |
The dms used in the analytic optative is distinguished from all other particles in that it can occur both by itself and with any person. Depending on the person, the denotation can be any sort of volitional, e.g., hortative, exhortative, imperative, etc. While Judezmo does not have the nonfactive/factive (dms/other complementizer) distinction of the other Balkan languages (Rmi te/kaj, Aro s(i)/cã, trã, Mac da/deka, oti, Alb të/që, se, Grk να/ότι, etc., see §7.7.2.1.3.1), the attachment of the complementizer ke to the verb, rather than occurring in clause-initial position, replicates the Balkan usage, as in: En ganedén ke esté ‘May s/he be in paradise’ (lit., ‘in paradise comp be.sbjv.3sg,’ Ammann & van der Auwera 2004: 306, from César Montoliu p.c.).
This thus represents a good example of a calqued grammatical syntactic change without copying material.Footnote 338
In Balkan Slavic and some (but not all) Romani dialects in contact with it, a choice of tense after the dms has the effect of softening or strengthening the volitionality. With the present tense, the effect is one of attenuation, whereas with the perfect the effect is one of emphasis, even threat, as in the following synonymous examples from these languages. Similar usages are attested in the Aromanian of Turia (Grk Krania) given in (6.154de) from Sobolev 2005a: 192. Sobolev makes the point that the imperative usage in (6.154d) is not acceptable in Greek, but the jussive usage in (6.154e) is acceptable in both languages:
a.
Da ne sum te videl pokraj nea (Mac) dms neg be.aux.prs.1sg you.acc see.lf beside her b.
Da ne săm te vidjal s nea (Blg) dms neg be.aux.prs.1sg you.acc see.lf beside her c.
Te na dikhlum tut džika late (Rmi) dms neg see.1sg.aor you.acc near her.loc ‘Let me not have seen you (= I’d better not see you) near her.’ d.
ś ti ai skulatã isĭa! (Aro) dms refl.2sg have.2sg.prs get.up.ptcp.f right.now ‘Get up right away!’ Cf.: *να έχεις σηκωθεί αμέσως! (Grk) dms have.2sg.prs get.up.nact.prf right.now e.
kãn vai γ’inu śi ai (adratã) etmã fãiĭa! (Aro) when fut come.1sg.prs dms have.2sg prepared.ptcp.f food.def Cf.: Όταν έρθω, να έχεις ετοιμάσει το φαγητό (Grk) when come.1sg.prs.pfv dms have.2sg prepared.prf the food ‘When I come, let the food be ready (lit., ‘let you have prepared the food’)
In (6.154c), Romani uses an aorist as the equivalent of the old perfect in Macedonian, which itself has an optative nuance. For (6.154d), the Macedonian equivalent would be almost identical, viz. da si stanal odma! ‘Get up right now!’ (lit., ‘dms be.2sg.prs get.up.lf immediately’), the only difference being that the Aromanian has a reflexive that is marked for person and a ‘have’ perfect, while the Macedonian simply uses the old ‘be’ perfect.
Albanian, too, uses the perfect optative to refer to a situation that is not past as in (6.155) (from Newmark et al. 1982: 90; cf. also Buchholz & Fiedler 1987: 151):
(6.155)
E moj Parti të qofsha falë O voc Party you.acc be.opt.1sg thank.ptcp ‘O Party, be thou thanked!’ (lit., ‘may I have thanked you’)
We can include here constructions that could be labeled impersonal optatives of the type ‘if only … .’ These overlap with attenuated conditionals, discussed in §6.2.4.2.8, where they are, for the most part, treated, although the two are not completely isomorphic. In addition to the attenuated conditionals that function as impersonal optatives, we can note constructions like that in (6.156a) from Macedonian, which resembles the Arvanitika examples in (6.191–192) and corresponds to të mos + subjunctive in standard Albanian (6.156b):
a.
Samo da ne pukne only dms neg burst(= explode).3sg b.
Vetëm të mos plasë only dms mneg burst(= explode).3sg.sbjv c.
Μόνο να μην σπάσει only dms mneg break.3sg.pfv.prs ‘If only it won’t break out [referring to a possible war]’
Relevant discussion occurs elsewhere: see §6.2.5.5 on the use of imperfect and perfect admiratives with the dms, which are basically past subjunctive irreal conditionals, §7.6.3 and §7.4.1.2.2.2.2 regarding prohibitions and other types of negated modals, and §7.6.3.2 for discussion of tentative questions introduced in Balkan Slavic by da ne and in Albanian by të mos or a mos, Rmi te na, etc. For subordinated optatives, see §7.7.2.1.1.1.6. Here we can also note another unsubordinated optative, one that is used for questions (which, if 1pl, has a hortative usage) using dms + present, as noted for Romanian by Zafiu 2013a: 46, citing Sandfeld & Olsen 1936: 352, Să plece? ‘shall s/he go?,’ Unde să meargă ‘Where shall s/he go?’ which can be found in most Balkan languages, e.g., Mac Da (si) odime? ‘Shall we go?,’ Što da pravam? ‘What shall I do?,’ Rmi So te kerav? ‘idem,’ Alb Çka të bëj? ‘idem,’ Grk Να πληρώσω τώρα; ‘Shall I pay now?,’ etc. In Turkish, the interrogative particle mI would normally be required except with an interrogative (WH) word, in which case Ne yapayım, with the optative, would be the translation of the Romani, Albanian, and Macedonian given above.
6.2.4.3.3 Hortatives
Balkan Slavic, Greek, Albanian, Romani, and Balkan Romance all have dedicated hortative particles or constructions. Of these Grk ας and BSl neka are hortative particles.Footnote 339 BSl neka can be followed by da in Bulgarian as well as in archaic or poetic registers of Macedonian (see also §§7.8.1.1, 7.8.1.2). The Balkan Slavic particle is limited to the 1st and 3rd persons, but Grk ας can occur with any person (Holton et al. 1997: 420), and, like the dms να but unlike αν ‘if,’ as noted in §6.2.2.3.3, it takes the modal negator μη(ν), as in (6.157):
(6.157)
Ας μην του το (έ)λεγες hort mneg him.gen it.acc tell.impf.2sg ‘You shouldn’t have told him it.’
Most Romani dialects use the lexical form muk ‘let,’ an imperative, + te as a 1st/3rd person hortative (Boretzky 1994: 147; Boretzky et al. 2008: 31). Some Romani dialects in contact with Balkan Slavic also use neka or ne (Petrovski & Veličkovski 1998: s.v.; Boretzky et al. 2008: 31), and some have me[k], which appears to result from the influence of the Slavic particle on the Romani (Boretzky et al. 2008: 64). Here it is worth noting that in some Romani dialects in Thrace and adjacent regions (and also Crimea), mek has eroded to me (or mi), i.e., a particle opaque in terms of its synchrony (Boretzky et al. 2008: Map 94). Albanian and Balkan Romance also have lexical ‘let’ as a hortative marker. In the case of Albanian, le + të (called ‘jussive’ by Buchholz & Fiedler 1987) can be used with all three persons, while Balkan Romance lasă (also Aro / Megl las or la s’) seems to be used like Balkan Slavic neka. Balkan Rοmance has hai să, using a hortative element (hai, cf. §4.3.4.3.3) along with the dms, e.g., Rmn hai să mergem ‘let’s go,’ Aro hai s-anchisim tora ‘let’s go now!’ (Carabaş 2015: 147). Some Aromanian dialects in contact with Greek have borrowed Greek ας (Beis 2000: 330–334).Footnote 340 All the Balkan languages can also use the ‘insubordinate’ dms (cf. §§6.2.4.2.8 and 6.2.4.3.2) as a hortative, e.g., Romani:
(6.158)
Te dika amen. dms we.see.prs we.acc ‘Let’s see one another.’
The use of a verb meaning ‘let’ in such contexts is not unusual, as the parallel with English let suggests. Thus, in general, aside from the parallel use of the dms as a hortative and the occasional examples of outright borrowing, Balkan hortatives can be seen as typologically general sorts of developments.
6.2.4.3.4 Imperatives and Prohibitives
Imperatives are subject to borrowing only as ERIC words, e.g., ela, (h)ajde, bujrum, etc., and these are covered in §4.3.10.2.2. As it happens, both Romani and Turkish use bare stems as 2sg imperatives, and so those Romani dialects with Turkish conjugation have the same rule for imperative formation regardless. Balkan Romance also uses bare stems for the imperative, and typologically such a rule is common and therefore not contact related. The imperative desinences of Balkan Slavic, Greek, and Judezmo are all inherited.Footnote 341
The differentiation of imperative and prohibitive (negative imperative) by means of a special negator involves an Indo-European inheritance in m-, Greek μη(ν), Albanian mos, and Romani ma, an innovated negator in Macedonian (and BCMS) nemoj and Bulgarian nedej, or a negated dms construction, as in Balkan Romance; see §§7.6.2, 7.6.3, 7.7.2.1.1.1.2.1, 7.7.2.1.1.1.2.2 on prohibitives and related phenomena. On the narrative use of imperatives, which is, arguably, a Balkanism, see §7.8.2.2.8.
6.2.4.3.4.1 Superordinate Aspect and Imperatives
In principle, imperatives are subject to the same aspectual expectations as indicatives: if the action expected is conceived of as single and/or terminative, it is perfective, and if not, it is imperfective. However, precisely because imperatives interact with such speech-event specific pragmatics as politeness and expectations, there is variation in aspectual assignment in those languages with superordinate aspectual distinctions. Thus, for example, in Slavic in general, in positive commands, the perfective imperative is normal, whereas in negative commands, the imperfective is normal. Under most circumstances, an affirmative imperfective imperative for a single act has a sense of ‘right now!’ and is therefore peremptory or impolite, whereas, in Macedonian, a negative perfective imperative has the nuance of a threat or a warning (in most Slavic languages, negative perfective imperatives are ungrammatical). Certain verbs in individual languages behave idiosyncratically and the affirmative imperfective imperative is normal, e.g., Russ sadites’ ‘sit down [polite],’ spite spokojno ‘sleep well [lit., peacefully],’ and Blg zapovjadajte, all imperfective but Mac povelete ‘please do …, etc.’(pfv), both BSl forms calqued on Trk Buyrunuz ‘[at your] command! (2pl)’; relevant here also are Alb urdhëroni, Grk ορίστε (from ορίζω ‘command,’ somewhat uncommon now in that meaning though found in Medieval Greek, and occurring now in the meaning ‘define’), Aro oriste, buiurun, buiurum, etc. (see also §4.3.10.1.2.2). The association of negative imperative with imperfective aspect does not appear firmly established in OCS, where both aspects can occur freely, although imperfectives are about three times as common as perfectives (Večerka 1996: 70). In Modern Greek, according to Benacchio 2013, the perfective imperative with the modal negator μη(ν) is less common than the imperfective imperative, and in at least some contexts the positive imperfective imperative has the same nuance of impatience as in Slavic. Given the antiquity of the Greek usage and the commonality among Slavic languages, albeit with some language- and subgroup-specific variation (Benacchio 2013), the similarities are probably typological rather than areal.
6.2.4.3.4.2 Number and Imperatives
The pragmatic feature of ‘politeness’ plays a role in number assignment in imperatives, but in many of the Balkan languages, this is connected with the relatively recent incorporation of pragmatic politeness into the plural, the well-known T/V distinction of Brown & Gilman 1960 and Friedrich 1972; see §6.1.4.3 for some details. This usage means that in many contexts the Balkan Slavic 2nd person plural affirmative imperfective imperative can only be used when addressing more than one person informally rather than just a single person, as its peremptory nuance makes it incompatible with polite singular reference.
In some of the languages, instances of imperative-like plural marking can be found on forms that ostensibly are nonimperatives. For instance, in Albanian, Newmark et al. 1982: 324 note that although only “rarely encountered in the [standard] spoken language, and still more rarely in the written language,” the 2nd person plural ending -ni can attach to various greetings and interjections in order to indicate that they are “addressed to a plural ‘you’ or politely to a singular ‘you’,” e.g., mirëditani ‘good day,’ mirëmëngjesni ‘good morning,’ and tungjatjetani ‘hello,’ forcani ‘heave-ho!’ (cf. forca ‘powers’), o burrani ‘onward!’ (with vocative particle o; cf. burra ‘men’), among others. The interjections implicitly have “a kind of imperative force” (Newmark et al., idem), so there is a certain synchronic rationale to the use of imperatival -ni with them, and even the greetings can be similarly construed (cf. the English leave-taking formula Have a nice day! with its overt imperative verb). This usage could be a carry-over from -ni’s origin in a free adverbial form *nū ‘now’ (Rasmussen 1985; Joseph 2011c), as it suggests a wider distributional range – such as a free adverb would have – for the element that developed into -ni; moreover, such an origin for -ni is consistent with the fact that weak object pronouns (clitics) can come between the imperatival root and -ni (see §7.5.3), under the reasonable assumptions that the free form that ultimately gave suffixal -ni could be positioned after the imperative-plus-object pronoun combination and that univerbation occurred somewhat later.
Moreover, in this regard, the exhortative exclamation discussed in §4.3.4.3.3 bears repeating here. (H)ajde ‘c’mon; gw’an; let’s go,’ although not an imperative etymologically, but rather most likely built from a Turkish interjection (see also §3.1), nonetheless is treated grammatically as if it were a singular imperative. That is, it has plural imperatival forms, with regular personal endings, for some languages both 1st person plural as well as 2nd person plural (Jašar-Nasteva 2001: 122):
a.
Romanian: (h)aideţi (2pl), (h)aidem (1pl) b.
Macedonian: ajdete (2pl) (but not *ajdeme), also shortened aj, ajte c.
Albanian: hajdeni (2pl) d.
Turkish: ha but also aydın, aydınız
In a somewhat parallel manner, Greek has an ostensible plural νάτε ‘here it is; take!’ built on the presentational element νά.Footnote 342 Note also Macedonian bujrumte from bujrum ‘please!, help yourself!, do come in!, what can I do for you?’ (from Turkish buyurun ‘command!,’ for which poveli is the Macedonian translated usage, cf. §4.3.10.2.2 and §6.2.4.3.4.1). Macedonian also has amante from the expression aman ‘mercy!’ (cf. §4.3.4.3.1); cf. also Turkish amanın (Jašar-Nasteva 2001: 123).
It is important to note that Slavic languages outside of Balkan Slavic show similar forms. BCMS, for instance, has (2pl) hajdete, (1pl) hajdemo, and Russian has pojdëmte ‘let’s go!,’ with 2pl -te added onto an inflected 1pl form pojdëm.Footnote 343 Such developments suggest a naturalness to the ostensible pluralization seen in Alb hajdeni (etc.) that could argue against contact being involved in these Balkan facts, unless Turkish or Slavic calques were the source for Greek and Albanian.
6.2.5 Evidentiality, an Overview
In a Balkan context, evidentiality (inferential, distance, mode of indirect narration, indirective, status, French médiatif, etc.) is a grammatical category encoding the speaker’s evaluation of the narrated event, often, but not always, predicated upon the nature of the available evidence. Evidentials can be of two types: confirmative (vouched for, ‘witnessed’) and nonconfirmative (not vouched for, ‘reported,’ ‘inferential’ etc.). The nonconfirmative can be felicitous (neutral report or inference) or infelicitous, in which latter case the nonconfirmative expresses either acceptance of a previously unexpected state of affairs (i.e., surprise, admirativity sensu stricto) or rejection of a previous statement (i.e., sarcasm, dubitativity). The opposition confirmative/nonconfirmative was already encoded in the Turkic simple past in -DI (confirmative) and the perfect participle in -mIş (nonconfirmative) at the time of the earliest monuments in the eighth century (Tekin 1968; Erdal 2004). In Balkan Slavic, the old synthetic pasts (aorist and imperfect) became markedly confirmative (this same meaning is also sometimes claimed for Torlak BCMS; Samilov 1957). By contrast, the old perfect using the resultative participle in -l has become an unmarked past (as it has throughout Slavic), with a chief contextual variant meaning of ‘nonconfirmative,’ and new tenses formed with the descendant of the resultative participle are markedly nonconfirmative. In Albanian, the inverted perfect (participle + ‘have’) has fused into a marked nonconfirmative present paradigm called ‘admirative,’Footnote 344 which can then function as an auxiliary to form analytic past tenses.Footnote 345 The Frasheriote Aromanian dialect of Bela di Suprã (Mac Gorna Belica) has reinterpreted the 3sg.prs Albanian admirative marker as an admirative suffix, which it adds to a masculine plural imperfect participial base to form a new admirative. Meglenoromanian also uses an inverted perfect (participle + auxiliary) for a nonconfirmative evidential, but, like the Balkan Slavic and unlike the Albanian, the Meglenoromanian evidential is a past tense. The Istanbul dialect of Judezmo has calqued the Turkis mIş-past by using a pluperfect in contexts where Spanish would not. The Romanian modul prezumtiv ‘presumptive mood’ uses dms, future, and conditional markers to encode the kind of evidential meanings associated with marked nonconfirmatives, and the Bulgarian dialect of Novo Selo in the Vidin region has developed a nonconfirmative evidential from a future marker. Several Romani dialects have developed various evidential strategies independently of one another. The Balkan developments all took place during the Ottoman period, and it is thus reasonable to assume that Turkish was the ultimate source for most if not all developments, although some language/dialects developed their categories via intermediaries and typological considerations are sometimes also relevant (cf. Friedman 2018b, 2019c, 2020b).
6.2.5.1 Turkish and Balkan Slavic
In both Turkish and Bulgarian, the same form can still be both a perfect participle and an evidential, whereas in Macedonian, the l-form is no longer a participle and can only be used to form analytic paradigms, the verbal adjective (descended from the old past passive participle but now derivable from any verb, both transitive and intransitive) having replaced any participial uses of the verbal l-form. The differences are illustrated in Table 6.28.
| Language | ‘time has [apparently/reportedly] passed’ | ‘passed time (= past tense)’ |
|---|---|---|
| Turkish | zaman geçmiş | geçmiş zaman |
| Bulgarian | vremeto minalo | minalo vreme |
| Macedonian | vremeto minalo | minato vreme |
The idea that the Turkish perfect in -mIş is marked as reported and the simple preterite in -DI is marked as witnessed appears already in Maḥmūd al-Kāšǧārī’s eleventh century Dīwān Luǧāt at-Turk [Compendium of the Turkic Dialects] (Dankoff 1982: 412). The relevant passage is worth quoting here:
The difference between these two forms [in -di and -miş] is that dāl yā’ [-di] on preterite verbs indicate that the action occurred in the presence of the speaker. The action was verified by its occurrence in his presence. For example, if someone says: BAR·DIY bardi the meaning is, “He went and I saw him go with my own eyes.” Mīm šīn [-miş], on the other hand, indicate that the action occurred in the absence of the speaker. Thus :: ’UL BAR·MIŠ ol barmiš “He went but I did not see him go”; :: ’UL KAL·MIŠ ol kälmiš “He came but I did not see him.” This is a general rule holding good for all preterite verbs, whether intransitive or not.
Although, as Johanson 1971 shows, and as Friedman 2000f demonstrates, the literal meanings of ‘witnessed’ and ‘reported’ are not obligatorily present in all uses of the the DI- and mIş-pasts, the basic opposition confirmative/nonconfirmative is sufficient for purposes of discussion here.
The situation in Balkan Slavic resembles that in Turkish, but Macedonian and Bulgarian have developed systems that are distinct from one another. According to Bulgarian normative grammar, the difference between the inherited perfect and the reported is the absence of the third person auxiliary in the latter and the claim that while the perfect can only be formed using an aorist-stem l-participle (which was indeed the situation in Common Slavic as illustrated by OCS), the reported can also be formed using an l-participle based on the imperfect stem. If this last condition were true, then the imperfect l-participle would never occur with a third person present auxiliary. Since in fact there are numerous instances of precisely such occurrences, even in normative usage, (e.g., Penčev 1967; Friedman 2004b), this last claim is demonstrably false. In standard Macedonian and the western dialects on which it is based, the third person present auxiliary in the ‘be’ perfect is consistently absent. (See Grickat 1954 on auxiliary omission in the former Serbo-Croatian; cf. also R. Greenberg 2000 on southern Montenegrin.) Macedonian, however, has developed a new perfect using the auxiliary ‘have’ and the verbal adjective (see §6.2.3.2.1), leaving ‘nonconfirmative’ as the chief contextual (but not the only) meaning of the old ‘be’ perfect. Moreover, both Bulgarian and Macedonian have developed new pluperfects using l-forms of the auxiliaries ‘be’ and ‘have’, respectively, which are markedly nonconfirmative.
As both Aronson 1967 and Friedman 2014b: 17–72, 2012a have argued, in both Bulgarian and Macedonian the synthetic past is marked for personal confirmation. By contrast, the analytic past formed with the l-form is not so marked, which favors a nonconfirmative interpretation such as report, inference, or disbelief.Footnote 346
The suggestion that what is called here nonconfirmativity was calqued into Bulgarian from Turkish dates back to Conev 1910/11. While the fact that in both Balkan Slavic and Turkish the so-called witnessed past (the DI-past in Turkish, the synthetic aorist and imperfect in Balkan Slavic), which is in fact not always literally witnessed but rather confirmative, must refer to a past action is widely accepted, the deployment of nonconfirmative forms is more complex.Footnote 347 It is often claimed that the so-called reported “mood” neutralizes tense (starting with Andrejčin [Andrejczin] 1938; see Friedman 2002b). In this sub-section, therefore, we have two tasks: (1) to determine whether contact with Turkish is involved in the development of Balkan Slavic evidentiality and (2) to determine whether these forms are indeed tenseless.
The first question was asked at least as early as van Wijk 1933: 243, who cites as possible evidence contrasting uses of the perfect and aorist to evaluate reports in the aorist in the oldest Slavic Paterikon. At issue however, is not the “reported” perfect, but rather the confirmative aorist (as a conclusion from reported evidence). The superficial formal resemblance between the omission of the third person auxiliary in the Bulgarian past indefinite and the addition of the Turkish copulative confirmative marker in -DIr is unquestionably not a matter of surface calquing, but rather of chance resemblance (Friedman 1978; Fielder 1999), especially given the fact that the third person auxiliary never occurs in the past indefinite in Macedonian and is the first lost elsewhere in Slavic, while -DIr can attach to any person and many tenses. While the concentration of confirmative meaning in the Balkan Slavic past definite could have begun prior to contact with Turkish, the developments as we see them today are clearly modern (in the broad historical sense, i.e., beginning circa the fifteenth century), i.e., precisely when Balkan Slavic was in intimate contact with Turkish, where the distinctions occur in the oldest documents (eighth century CE).
At issue in the question of tensed versus tenseless is whether the so-called “reported” can claim the kind of autonomous status as a separate category that is clearly the case in some languages of North and South America (Friedman 2018b). In the case of Turkish and Balkan Slavic, instances such as example (6.160), which can be interpreted as perfect/past as in (a), reported/inferential as in (b), admirative as in (c) and dubitative as in (d), are regularly cited as examples of tense neutralization, although (c) also presents problems for a theory that relies on reportedness or any related concept:
(6.160)
Toj bil dobar čovek (Balkan Slavic) he be.prf good person İyi adam imiş (Trk) good person be.prf a.
a. He has been/used to be a good man. (perfect/past) b.
They say/said/Apparently he is/was a good man. (reported/inferential) c.
It turns out he is/was a good man! (admirative) d.
Oh, sure, he is/was a good man (and cows can fly)! (dubitative)
Meaning (6.160a) is a straightforward perfect, although in context it can also refer to a past state of affairs whose present relevance is to the conversation at hand rather than something actually continuing into the present. (Note that Standard Bulgarian would require a 3.sg.prs.cop e ‘is’ before bil for the perfect reading, but in many dialects, and under certain pragmatic circumstances even sometimes in literature, the copula is omitted; cf. Friedman 2002b.) In every instance, however, there is some form of past reference, even when the best English translation is a present tense. In (6.160b) there is always some previous report or pre-existing state of affairs; admirative usage in Balkan Slavic and Turkish always refers to a pre-existing state of affairs, and dubitative usage always refers to an explicit or implicit previous statement or assumption. As argued in Friedman 1977b, 2000b, a general meaning (Gesamtbedeutung) of nonconfirmative can account for all usages insofar as they refer to infelicitous (admirative) and felicitous (dubitative, reported/inferential) deployments of the concept ‘nonconfirmative,’ itself a result of the development of marking for confirmativity in the -DI and synthetic pasts of Turkish and Balkan Slavic, respectively. The evidence for this comes from interrogatives as in (6.161):
a.
Kade bil majstorot? (Mac) ‘Where *is/was the boss?’ Where was.ncnfv boss.def b.
Usta neredeymiş? (Trk) ‘Where *is/was the boss?’ boss where.be.ncnfv c.
Ku qenka mjeshtri? (Alb) ‘Where is/*was the boss?’ where was.adm.3sg boss.def
The context for (6.161) is the following: a man walks into a barber shop and sees the barber’s apprentice but is surprised that the barber himself is not in his shop. At this moment, he requests information about the current whereabouts of the boss. The question does not refer to a previous state of affairs, and therefore can use the present admirative in Albanian (see §6.2.5.5) but not the Balkan Slavic or Turkish nonconfirmative. If, however, the customer were to ask, e.g., in Turkish, “Usta nerede?” (‘Where is the boss?’, lit., ‘boss where’) and the apprentice were to answer that he didn’t know, that he wasn’t around, that he wasn’t at home, etc., and the exasperated customer did not believe him, he could then exclaim: “Iyi be, usta neredeymiş?!” ‘OK, then, where is the boss?!’ but this quotation would be an exclamation of sarcastic exasperation at the apprentice’s previous responses rather than a genuine question. The same holds true for the Balkan Slavic equivalent. Moreover, in the contexts of meanings (6.160b–d) Albanian could use not only a present admirative, but any of its past admiratives (see Table 6.30) in exactly the same set of meanings.
Forms such as Bulgarian bil napravil and Macedonian imal napraveno ‘he reportedly/supposedly/apparently has/had done [it],’ which use the old perfect of ‘be’ or ‘have’ as an auxiliary and which date from the late medieval and early modern periods, respectively, are new paradigms created after the synthetic past became a marked confirmative. These marked nonconfirmatives begin as pluperfect constructions and then, in southwestern Macedonia the new perfect pushes the old perfect into marked nonconfirmativity and, ultimately, almost oblivion (Friedman 1986a, 1988b, 2018b; see now also Makartsev 2013a, 2013b).Footnote 348 See also §6.2.5.6.
6.2.5.2 Judezmo
The Judezmo of Istanbul uses the pluperfect as a calque on the Turkish use of its perfect marker -mIş in its function as a nonconfirmative, reported, or unwitnessed past (cf. Friedman 2012d). In examples (6.162a) and (6.163), cited in Varol 2001: 91,Footnote 349 the pluperfects aviya entrado ‘he had entered’ and s’aviya etcho ‘he had become,’ would not be grammatical in Standard Spanish. In example (6.162a), the effect is to calque the Turkish perfect in its unwitnessed meaning – illustrated in example (6.162b) – while in example (6.163) the effect is one of reportedness, which is another meaning conveyed by the nonconfirmative use of the perfect in Turkish:
a.
Kuando estavan en l’ Amérika, les aviya entrado ladron when they.were.impf in the America them.dat had.impf enter.pst.ptcp thief ‘When they were in America [i.e., absent], a thief (apparently) broke into their house.’ b.
onlar yok-ken, hırsız gir-miş they not.exist-while thief enter-prf ‘While they weren’t there, a thief entered’
(6.163)
Dos ermanos eran, uno salyó doctor two brothers were.3pl.impf one he.became.pst doctor salyó dahilkiye después s’ aviya etcho became.pst internist afterwards refl had.impf made.pst.ptcp dişçi, el otro doctor de bebés dentist the other doctor of babies ‘There were two brothers, one became a dentist and the other became an internist, afterwards he became (lit., ‘had become’) a pediatrician.’
This kind of evidential taxis is attested in other forms of Spanish influenced by languages with evidential systems or usages, e.g., in the Spanish of Peru, where the pluperfect is used to render evidential effects, much as in the examples cited here, owing to the substratal influence of Quechua’s evidential system (p.c. from Carlos Arregi, Peter Muysken, and Dan Slobin; cf. Friedman 2018b: 134). Similarly, the Cilician Arabic of the Çukurova region of southeastern Turkey uses participles as a calque on the Turkish mIş-past in its evidential uses (Procházka-Easl & Procházka 2018).
6.2.5.3 Romani
Evidential strategies occur in a variety of Romani dialects, both Balkan and non-Balkan, although all but one are from the Balkans.Footnote 350 In the Balkans, evidential marking is always by means of particles (except in the case of borrowed Turkish conjugation, cf. Friedman 2013b and §6.2.1.1.1). Thus, for example, in Futadži Romani (Haskovo, SE Bulgaria), the past in -mIş is used with Turkish verbs in contexts where it would be expected in Turkish, while with native Romani past tense verbs in the same contexts, the particle berim is normally placed immediately after the verb or its pronominal object if that follows the verb.Footnote 351 Example (6.164), from Ivanov 2000: 39, illustrates this effect of Turkish on Futadži Romani not attested elsewhere.
(6.164)
Oda kana dikljas la berim don-muš pe taneste that.m when see.3sg.pst her.acc berim freeze-prf in place.loc ‘When he saw her [berim], he froze in his tracks [lit., ‘place’].’
Boretzky 2018: 37–45 considers berim to have come from Turkish belli ‘sure, known,’ but the postposition berin ‘according to this’ is another possible candidate, as it is phonologically closer and semantically could carry a notion of secondary attribution appropriate to the nonconfirmative meanings of the Turkish perfect.Footnote 352 The Futadži phenomenon is unique in that it is a native development co-occurring with an integrated foreign (Turkish) category.
Kostov 1973: 107–108 and Igla 2004, 2006 describe a different nonconfirmative evidential in the (Balkan II) dialect of Romani spoken in Sliven in eastern Bulgaria. The evidential strategy is realized by the suffixing of the particle li, transparently of Slavic origin, to a finite verb to render the nonconfirmative meanings as in Bulgarian or Turkish.Footnote 353 Examples (6.165a) and (6.166a), from Kostov 1973: 107–108, are from a folktale and conversation, respectively, the first with an imperfect, the second with an aorist.Footnote 354 Examples (6.165b) and (6.166b) give the Bulgarian equivalents. Igla 2004, 2006 supplements and updates the data in Kostov 1963, 1973 and Igla 2006: 60 supplies an example of emotive usage with doubled li, given here as (6.167):
a.
u thagar uslovijas-li pis lenca (Rmi) the king come.to.terms.impf.sg-li intr them.ins b.
carjat se uslovil s tjah (Blg) the.king intr come.to.terms.lf.m with them ‘The king came to terms with them’
a.
oda vakerjas mangi, či tu phirsas-li. (Rmi) he say.aor.3sg me.dat that you go.impf.2sg-li b.
To mi kaza če ti si xodel. (Blg) he me.dat say.aor.3sg that you be.2sg went.lf.m ‘He told me that you were/are going.’
(6.167)
mu papus kana avilas-li piski – ne habi isja-lili, my grandfather when come.aor.3sg-li self.dat not food was-li+li nikačhi – u kažaja hana-li sourru (Rmi) nothing the children eat.aor.3pl-li everything ‘When my grandfather came home, there was no food, nothing. The children had eaten it all.’
Example (6.167) is taken from an anecdote narrated by a speaker about when his father was a boy. The narrator’s father and grandfather had gone out to work while the other children remained at home. The narrator’s grandmother made food, but the children at home ate it all up, and so when the narrator’s grandfather and father came home, there was nothing for them to eat. All of the verb forms are past tenses marked with li, but the form isja-lili, from isja ‘was,’ has a double marking of -li. Igla compares this to the so-called emphatic reported of Bulgarian (the Bulgarian equivalent would be nemalo bilo), which Aronson 1967 labels ‘dubitative.’ Here, however, the effect is one of surprise or shock rather than pure disbelief. Igla analyzes this as if it were the report of a report, i.e., the grandmother’s original presumed statement that there was no food, as reported by the father to the narrator. However, in this context the doubled li could also represent a meta-commentary of the narrator with admirative meaning, i.e., the narrator is rendering the surprise felt by his father and grandfather.
Igla 2006: 61 also supplies an example of clear admirative usage, given here as (6.168), where the speaker expresses surprise at a discovery:
(6.168)
O Devla, ta oda mandar da butrašadi isja-li O God.voc and they me.abl and more.scared be.impf-li ‘Oh my God, they’re more timid than I am!’
The exclamation is made by a rabbit upon seeing frogs jumping into the Danube. Igla glosses isja as present, which is indeed the effect, but the form itself is imperfect. This is classic admirative usage of a past tense form to express surprise at the discovery of a pre-existing state, with li in Sliven Romani being essential to this interpretation.
Although the particle li is clearly of Slavic origin, Kostov 1963: 123, 132–33, 1973: 107–108 and Igla 2004, 2006 give different explanations of its precise source. According to Kostov, the Sliven Romani use of li is based on the -l that is used to form the Bulgarian past participle that serves as the basis of the perfect and of evidential strategies and paradigms. Igla 2006: 56, however, conjectures that its source is an expressive use of the interrogative particle li.
As Friedman 2013c, 2019b argues, data from Kriva Palanka Arli (KPA) and Skopje Barutči Arli – as well as typological parallels – support Igla’s suggestion. In these Macedonian Romani dialects, a borrowed interrogative particle (Mac li in KPA and Trk mi in Skopje Barutči) can be used to mark a declarative dubitative, as in examples (6.169abc). The context for the example is a telephone conversation. Someone telephones the speaker of (6.169) and claims to be calling from America. The speaker knows that this is not possible, and responds with (6.169a) in Macedonian, (6.169b) in Kriva Palanka Arli Romani, and (6.169c) in Skopje Barutči Romani.Footnote 355
a.
Abe ti si bil vo Amerika. Lažeš! (Mac) voc you be.prs.2sg be.lf.m in America lie.prs.2sg b.
Abe tu hinjan li t-i Amerika. Hohavea! (Kriva Palanka Arli Rmi) voc you are li in America lie.prs.2sg c.
Abe tu injan mi ki Amerika. Hohavea! (Skopje Barutči Rmi) voc you are q in America. lie.prs.2sg ‘Oh sure, you are/were in America! You’re lying!’
The Macedonian version is ambiguous between a present and past interpretation, as would be its Bulgarian equivalent, but thе Romani usage is with a present tense. The Barutči use of the Turkish interrogative particle mi precisely where Kriva Palanka Arli has li is a clear indication that a shift from interrogative marking to dubitative marking is the basis of this type of dubitative usage. This in turn suggests the route by which the Sliven Romani usage arose. The connection of interrogation with nonconfirmation is clear insofar as interrogation is, felicitously, a request for information about something the speaker does not know and therefore cannot confirm. Used infelicitously, however, i.e., sarcastically, interrogation becomes an expression of active nonconfirmation, i.e., disbelief. From that point, the way is then open to dubitative usage that in turn can lead to neutral nonconfirmation based on report or inference as well as expressions of felicitous surprise such as that found in Sliven Romani.
Relevant here is the fact that expressive, i.e., noninterrogative, li also occurs in BCMS, as in (6.170), which was recorded in the early 1990s. The sentence has the structure of a rhetorical question marked with li, but is in fact an exclamation expressing the speaker’s enraged shock at the addressee’s failure to comply with his urgent request:
(6.170)
Pa pička vam materina, jesam li rekao da mi pošaljete vod and cunt you.dat mother’s am q said that me.dat send.prs.2pl platoon vojnika sad, jebem li vam hljeb i boga! soldier.gen.pl now fuck.prs.1sg q you.dat bread and god ‘And your mother’s cunt, I said li to send me a platoon of soldiers now, I fuck li your bread and god!’ (more idiomatically: ‘You stupid mother-fuckers, I told you to send me a platoon now, God fucking damn you to hell!’)
Such usage is identified as a Serbism by native speakers of Macedonian. However, like the preposition u ‘in, at’ (standard vo) or the pronoun mi ‘we’ (standard nie), it occurs in urban Skopje speech and could represent a northern Macedonian dialectism as well as a Serbism.Footnote 356 This would be consistent with the Kriva Palanka and Barutči developments, as they are both located in (or at least in contact with in the case of Barutči) the northern dialect region.
This usage of li can be argued to be at the intersection of admirativity and dubitativity: the speaker is expressing his enraged shock and disbelief at the interlocutor’s failure to comply with his request. The parallel with the Macedonian Romani dubitative usage is the fact that in colloquial conversation the interrogative particle can have noninterrogative uses that emphasize the speaker’s relationship to what is being said. As argued elsewhere (Friedman 2014b, 2012a), the meanings of admirativity and dubitativity are separated only by presuppositional felicity conditions.
A close typological parallel provided by the use of the interrogative marker in combination with a negative optative in Turkic languages functions as an expression of the unexpected, i.e., as a kind of admirative, as in the following example from Nasilov et al. 2001: 218 in Tatar (6.171a) and its Turkish translation (6.171b):
a.
Kič belän Färid kilep ker-mä-sen-me? evening with F. coming enter-neg-opt-q b.
Akşamleyin Ferit gelip gir-me-sin mi? evening.with F. coming enter-neg-opt q ‘Unexpectedly Farid showed up this evening.’
Here Tatar kil- and Turkish gel-, both meaning ‘come,’ function as converbs expressing unexpectedness when used with the negative interrogative optative. Kazakh, Kirghiz, and Uzbek have similar usages (Kagan Arik, p.c.). Example (6.171) could also be translated: ‘In the evening, who should show up but Farid!’ This kind of rhetorical question expressing surprise is similar to the English wouldn’t you know it, which is not a question. Moreover, the use of a lexical motion verb to express unexpected or sudden action is similar to the English usage he went and did it. While these examples are also more admirative than dubitative, the relationship of surprise to doubt discussed above is relevant to explicating the path by which an interrogative marker comes to be used in evidential strategies.
The connections of perfects and quotative particles, on the one hand, to evidential strategies, on the other, are well attested in many languages, but the transition from interrogative marker to evidential strategy marker appears to be uncommon. The Turkic usage illustrated by example (6.171) shows the kind of association of interrogation with the unexpected that can in turn be connected to expressions of surprise at something discovered or doubt of something said by another. The BCMS usage of li in (6.170) shows how an interrogative marker can form a rhetorical question whose effect is in fact the expression of surprise and dismay at another speaker’s actions (or, in this concrete case, failure to act). The copying of interrogative markers from contact languages into Romani is a well-attested phenomenon, but their expressive use to mark dubitativity is a specific, Romani-internal development. The use of interrogative particles to mark dubitativity in certain Macedonian Arli dialects as illustrated by (6.168) and (6.169) is an indication of how the evidential strategy attested in Sliven Romani in (6.164)–(6.168) could have arisen independently, albeit under similar conditions. These facts demonstrate how typological (universal) and areal (contact) explanations can be used together in a nuanced fashion, and without conflation, to account for language change. The connection between interrogation and dubitativity is typological, but the development of interrogative markers into dubitative or evidential markers in Romani dialects in North Macedonia and Bulgaria occurred precisely in contact with languages that already had evidential strategies in their grammars.
In some dialects of Romani in eastern Bulgaria, the Turkish morpheme -mIş in its high back rounded variant -muš accompanied by the 1sg marker -um, acts as an evidential particle, as in (6.172)–(6.174) from Imrenčevo, Veliki Preslav commune, Šumen district, in northeastern Bulgaria. All data in Imrenčevo Romani are from Kyuchukov 2012. The particle expresses the full range of the admirative complex, i.e., neutral report (6.172), surprise (6.173), and disbelief (6.174):
(6.172)
Raci o kyzaj peris-mušum avri. last.night def.m. child fall.down.aor-mušum outside ‘Last night the child (apparently) fell down outside.’
(6.173)
Tu but gozaver sijan-mušum. you very clever are-mušum ‘You are very clever! (to my surprise).’
(6.174)
Xaj tu but gozaver sijan-mušum. supposedly you very clever are-mušum ‘You are very clever, indeed! (what rot!).’
In the tale Tilči Bej (Cech et al. 2009: 214–239), which can be translated ‘Lord Fox’ or ‘Lord Vixen’ and is essentially a fairy tale of the type known to Anglophone readers as Puss in Boots, there are three occurrences of the Turkish miş-past copula, 3rd person imiš, and, as it turns out, both imiš and imišim (1st sg) occur as evidential markers in Skopje Romani dialects. The tale was recorded from one of the oldest speakers in Cech et al. 2009, Uka Muarem, born in Skopje in 1896. There are seven identifiable Arli dialects spoken in Skopje (Friedman 2017a), and although the collector of the tale gave no indication of which group the speaker came from, both the general dialectal evidence as represented in the tale (despite occasional dialectal mixing) and the judgments of modern speakers indicate that the dialect is Gilanli. The three occurrences of imiš in Tilči Bej, are given here as (6.175–6.177):Footnote 357
(6.175)
Avol jek lisica. Oj imiš, oj imiš bego, odoja lisica. come.3sg.prs one fox she imiš she imiš lord that fox Tilči Bej vakeren. T. B. say.3pl.prs ‘A fox has come. This fox, she says, she says she’s a lord, that fox. They call her Tilči Bey.’
(6.176)
So ali, odma i puška te khuvol la: what come.pst.3sg.f immediately def gun dms kick.3sg.prs her So mange, po kaj sijan imiš tu? what want.3sg.prs and where are imiš you ‘As soon as she came, he stuck a gun in her face: “What do you want? Where the hell have you been?”’
(6.177)
Garav to šero. Ale gotovo! Ma te dikhen tu imiš. hide.impv your head come.3pl.pst all.done mneg dms see.3sg.prs you.acc imiš ‘Hide your head. They have come, it’s done. Don’t let them see you!’
Example (6.175) is a straightforward reported evidential usage, but what is noteworthy about it is the fact that imiš here functions as a finite verb rather than as a particle. The Romani dialects of Skopje do not have Turkish conjugation (cf. Friedman 2013a), and this type of usage seemed strange to speakers born after 1970 or so, for whom Turkish is a language their parents and grandparents know/knew but they do not. Thus, the usage in both (6.176) and (6.177) is emphatic, albeit not readily assignable as admirative or dubitative. Topaanli speakers of the younger generations found the sentences strange, but they are generally consistent with the link between the admirative complex and emphatics discussed above.
Although the usage in Tilči Bej represents an older generation of Gilanli, the use of imiš as well as the 1st singular imišim is also known in Topaanli, even for the younger generations who no longer know Turkish, as seen in (6.178) and (6.179):
(6.178)
Geljum te bešav leste ko kher, ama i romni come.1sg.pst dms sit.1sg.prs him.loc in house but def wife vakerde imiš na bešela više adathe. say.3sg.pst imiš neg sit.3sg.prs more here ‘I went to stay at his house but his wife claimed that he doesn’t live there anymore’
(6.179)
hem maškar ko čhaja ine but, but šuži, odoborom ine šuži and among in girls was very very beautiful so.much was beautiful so, imišim kotar ka dikjhela našti ti jak te irane what imišim from.there fut see.3sg.prs not.able your eye dms return.2sg.prs ‘And among the girls there was a very, very beautiful one, she was so beautiful, apparently, if you saw her, you wouldn’t be able to take your eyes off her.’
Examples (6.178) and (6.179) illustrate the two kinds of uses of imiš[im] current among younger and middle-aged speakers of Topaanli Romani, namely dubitative (6.178) and neutral report (6.179). In (6.178), the effect of imiš is captured by the substitution of ‘claim’ for ‘said’ in the English translation. The speaker is explicitly refusing to endorse the wife’s statement and is casting doubt on it. Example (6.179), however, is a neutral report. Here the speaker is simply emphasizing that s/he was not there without casting any doubt on the report. It is interesting that precisely when the report is a neutral one, there is free variation between imiš and imišim, but when doubt is implied, then only imiš is used. Cf. the discussion on (6.247) in §6.2.5.10.
6.2.5.4 Meglenoromanian
The Meglenoromanian inverted perfect, which in Romanian is a stylistic variant of the perfect, has been reinterpreted as a nonconfirmative, with exactly the same complex of meanings of reportedness, surprise, and disbelief found in the nonconfirmative uses of the Turkish and BSl (old) perfect. Given the intensive contact with both Turkish and, especially, Macedonian, these languages are the likely sources of the semantic development. Thus, for example 1sg.prf of ‘see,’ ăm vizút, inverted perfect vizút-ăm, and the inverted perfect of ‘have,’ vutăm, can then function to form an analytic pluperfect vutăm vizút. Unlike the Romanian inverted perfect, where weak forms like the intransitive/reflexive s come between the participle and the auxiliary, e.g., dusu-s-a ‘he has gone’ (go.ptcp – anaptyctic u + intr + aux), in the Meglenoromanian nonconfirmative the weak element precedes the entire form si-turnat-ăṷ ‘he has [apparently] returned, etc.’ (Graur et al. 1966: 269; Atanasov 2002: 254). Table 6.29, based on Atanasov (1990), illustrates the forms.
| Perfect | Pluperfect | ||
|---|---|---|---|
| vizút-ăm | vizút-ăm | vutăm vizút | vut-ăm vizút |
| vizút-ăi̯ | vizút-ăț | vutăi vizút | vut-ăț vizút |
| vizút-ăṷ | vizút-ăṷ | vutăṷ vizút | vut-ăṷ vizút |
Examples (6.180–6.183) illustrate admirative, reported, and narrative uses of the Meglenoromanian evidential strategy (note the bolded forms and translations): (6.180ab) is an admirative given with its Macedonian equivalent, (6.181) is a neutral nonconfirmative past, and (6.182) is a neutral nonconfirmative pluperfect:
a.
ă bră, tu fost-ăi̯ mári om! (Megl) voc you been-have.2sg big man b.
a be, ti si bil golem čovek! (Mac) voc you be.2sg be.lf big person ‘Well, [I just learned that] you are an important person’ (Atanasov 1990: 221)
(6.181)
țirút-ăṷ si víă la noii̯ (Megl) wanted-have.3sg dms come.3sg to us ‘[Apparently] he wanted to come to our place.’ (Atanasov 1984: 528)
(6.182)
tu vút-ăi̯ măncát cǫ´n vinít-ăṷ i̯ăl (Megl) you had-have.3sg eaten when came-have.3sg he ‘[They say] you had eaten when he came.’ (Atanasov 1984: 528)
Finally, (6.183ab) are the beginning of a folk tale with its Macedonian translation.
a.
Ash vut-aṷ ună-ṷară ună mūmă trei̯ feati shi li trimes-aṷ self had-have.3sg one-time one mother three girls and them sent-have.3sg la spilarea la vāli. Cum spilat-ăṷ cṷola vinit-ăṷ ună to washing to river. as washed-have.3pl there came-have.3sg one meară di pri vali (Megl) apple from along river Si imala ednaš edna majka tri kjerki i gi ispratila self have.lf.f once one mother three daughters and them sent.lf.f na perenje na rekata. Kako što perele tamu, došla to washing to river.def as what they.washed.lf.pl there came.lf edna jabolka po rekata (Mac) one apple along river.def ‘Once upon a time a mother had three daughters and she sent them to do laundry at the river. As they were washing there an apple came [floating] down the river.’
This last example demonstrates that the Meglenoromanian nonconfirmative, as in Balkan Slavic and Turkish but unlike the Albanian and Gorna Belica Frasheriote Aromanian (GBFA) admiratives discussed in §6.2.5.6, can be used in extended narratives.
6.2.5.5 Albanian
The Albanian admirative is a marked nonconfirmative with a dedicated present tense in addition to a series of past paradigms (cf. discussion in §6.2.5.1). In this way, it differs sharply from Balkan Slavic, Turkish, Judezmo, and Meglenoromanian, all of which lack a formal difference between present and past interpretation of the relevant forms. Moreover, the Albanian admirative, like its GBFA equivalent (see §6.2.5.6), and unlike Balkan Slavic and Turkish, is not used in connected narratives. Examples of the relevant forms are given in Table 6.30.
| Paradigm | Nonadmirative | Admirative |
|---|---|---|
| Present | kam | paskam |
| Perfect | kam pasur | paskam pasur |
| Imperfect | kisha | paskësha |
| Im-Pluperfect | kisha pasur | paskësha pasur |
| Perfect-2 | kam pasë pasur | paskam pasë pasur |
| Pluperfect-2 (imperfect) | kisha pasë pasur | paskësha pasë pasur |
| Aorist | pata | – |
| Ao-Pluperfect | pata pasur | – |
| Pluperfect-2 (aorist) | pata pasë pasur | – |
As can be seen from Table 6.30, the present admirative is based, etymologically, on an inverted active perfect (historically, participle + auxiliary, but synchronically an inflectional paradigm), an imperfect derived from an inverted pluperfect (same principles as with the inverted perfect), and various analytic paradigms using the admiratives of ‘be’ and ‘have.’ Owing to the similarity in meaning of the Albanian admirative with the admirative (and dubitative and reportative) usages of the l-past in Balkan Slavic and the mIş-past in Turkish, a similarity noticed by Weigand 1923–24, 1925; Sandfeld 1930: 178, and others, earlier scholars equated the Albanian paradigms with the Balkan Slavic and Turkish usages. However, precisely when Balkan Slavic and Turkish can use their respective pasts with apparent present meaning, Albanian has the option of using any of its admirative forms regardless of tense. Thus, for example, the Albanian present admirative translation of example (6.160) in §6.2.5.1 would be ai qenka njeri i mirë (lit., ‘he be.adm.prs.3sg man good’). However, precisely in a context when Balkan Slavic and Turkish can use their respective pasts with an apparent present meaning (always referring to a newly discovered or otherwise reported previous state of affairs), Albanian can use a perfect (6.184a), pluperfect (6.184b) or imperfect (6.184c) admirative; the Bulgarian in (6.184) represents the original, from Konstantinov 1895, and the other versions are translations. Examples (6.184ab) are direct translations, whereas (6.184c) translates a similar sort of sentence (all from Friedman 2000f):
(6.184)
Brej, hepten magare bil tozi čovek. (Blg) voc total ass be.lf this man Brej, epten magare bil toj čovek! (Mac) voc total ass be.lf this man Vay anasını, bu herif hepten de eşekmiş be! (Trk) woe his.mother.acc this guy altogether also ass.miş voc a.
Bre! gomar i madh paska qënë ky njeri! (Alb) voc ass pc big have.adm.prf this man b.
Ore, fare gomar paskësh qënë ky njeri! (Alb) voc nothing ass have.adm.plu this man ‘What an ass that guy is!’ c.
Ama njerëz fare pa mend qënkëshin këta (Alb) but men nothing without thought be.adm.impf these ‘What fools these men are … ’
Another point worth making in connection with the Albanian admirative vis-à-vis Balkan Slavic and Turkish is the fact that in the respective translations of Konstantinov 1895, only 10 percent of the Albanian present admiratives corresponded to admirative usage in Balkan Slavic and Turkish, the remaining 90 percent were plain presents or otherwise noncorresponding (Friedman 1982b).
The semantic development of the admirative as it occurs in standard Albanian and many Albanian dialects today took place during the early modern period, i.e., when there was contact with Turkish. As Friedman 2010b has shown, in the oldest significant Albanian text (1555), the Missal of Gjon Buzuku (Çabej 1968), the inverted perfect functioned primarily as a conditional-optative (cf. also Sh. Demiraj 1971). Examples (6.185) and (6.186) are typical: here (and below – 6.187–6.189, 6.193, 6.195), we give them first from Buzuku (a) and then with translations into three modern Albanian versions: Bible 1872 from the late nineteenth century (b), Bible 1930 from between the two World Wars (c), and Bible 1980 in the modern post-1968 unified standard (d).Footnote 358
a.
Ju në klënëkishte të shekullit, shekulli t’ u dashke pori të vetë (Buzuku LXXXII) if be.inv.plu.2pl dms refl love.inv.prf b.
Ndẹ yễni prei bŏtẹs, bota dotẹ donte tẹ vĕtinẹ (Bible 1872) if be.2pl.prs fut.dms love.impf c.
Të ishitë prej botësë, bota dote donte të-sajnë (Bible 1930) dms be.impf.2pl fut.dms love.impf d.
Po të ishit të botës, bota do t’ i donte të vetët (Bible 1980) if dms be.impf.2pl fut dms them love.impf. pc self ‘If you were of the world, the world would love its own’ (John 15:19)
a.
E kȳ në mos ish ṇ Sinëzot, nukë të mujtëke me bām as ṇdonjë kafshë (Buzuku LVII) neg dms can.inv.prf b.
Ndẹ mos ište kǚi prei Perẹndisẹ nukẹ mŭndei me bẫmẹ asǵẫ. (Bible 1872) neg could.impf c.
Ndë mos ishte ky nga Perëndia, nukë dotë muntte të bënte asgjë (Bible 1930) neg fut.dms can.impf d.
Po të mos ishte ky njeri prej Hyjut, s’ dotë mund të bëhet asgjë! (Bible 1980) neg fut dms can ‘If this man were not from God, he would not be able to do anything.’ (John 9:33)
This type of expressive modal admirative usage also occurs in contemporary Albanian, as the following examples (6.187–6.189) show; all of these are counterfactuals involving the dms + impf or prf (i.e., past subjunctives). According to Sh. Demiraj 1976: 281, they are basically synonymous with their nonadmirative equivalents. Buchholz & Fiedler 1987: 158 observe that they are limited to a few authors, particularly Naim Frashëri and Jakov Xoxa, who are the sources for (6.187–6.189). Example (6.187) is an admirative imperfect subjunctive functioning as an attenuated conditional or optative (cf. §6.2.4.2.8), (6.188) is a similar ‘as if’ type clause (cf. §6.2.4.2.8), while (6.189) shows subjunctive perfect admiratives in the protasis of an irreal conditional (cf. §6.2.4.2.3):
(6.187)
Të paskësha vrap’n e veriut, të kishnja krahë pëllumbi dms impf.adm.1sg dms impf.1sg ‘If I had the speed of the north wind, if I had the wings of a dove … ’ (N. Frashëri = Sh. Demiraj 1971: 39; Buchholz & Fiedler 1987: 159)
(6.188)
(6.189)
Sikur e gjyshja të mos paskësh nxjerrë as.if dms mneg have.impf.adm.3sg stick.out.ptcp e të mos e paskësh thirrur and dms mneg him have.impf.adm.3sg call.ptcp kokën nga qerrja ‘If his grandmother had not actually stuck her head out of the cart and had actually called him, [who knows how long he would have remained like that, standing (in the midst of) the rain that had settled over the plain.]’ (Xoxa, cited in Sh. Demiraj 1976: 282; Newmark et al. 1982: 86)
For the most part, Arvanitika, like the Lab and Çam dialects, does not have admiratives. Liosis 2010, based on Reinhold’s 1855 material from Hydra, has found inverted pluperfects (which in standard Albanian would be an imperfect admirative) surviving as a counterfactual or attenuated conditional, as in (6.190)–(6.192):Footnote 359
(6.190)
t’i márrəkeshə, do i kéjə dms+them take.impf.adm.1sg fut them have.1sg.sbjv ‘If I had taken them I would have them’ (Liosis 2010: 186)
(6.191)
tə mos arrə´təkeshə dms mneg come.impf.adm.1sg ‘İf only I hadn’t come!’ (Liosis 2010: 187)
(6.192)
mos u ndódhəkej mneg intr happen.impf.adm.3sg ‘I wish he hadn’t been born’ (Liosis 2010: 187)
Even in Buzuku, however, the inverted perfect was also beginning to show some admirative functions, as seen in the three indicative examples: In modern terms, the three forms are present, imperfect, and perfect, respectively. The first example comes from the feast at Cana, where Jesus turns water into wine. The master of the feast, having tasted the miraculous wine, calls the bridegroom over and says to him: “Every man brings out the good wine first, and when people are drunk he brings out the inferior wine” (see above regarding (6.185) and Footnote footnote 311):
a.
e ti ruojtëkē venënë e mirë djerie tash. (Buzuku XXXVIII) keep. inv.prf.2sg b.
por ti ke rŭeitunẹ vễnẹn e mirẹ ḱṳš taštĭ (Bible 1872) have.2sg kept.ptcp c.
po ti ke ruajturë verën’ e-mirë gjer ndashti. (Bible 1930) have.2sg kept.ptcp d.
Ti paske ruajtur verën e mirë deri tani. (Bible 1980) have.adm.2sg kept.ptcp ‘but you have kept the good wine until now.’ (John 2:10)
Demiraj argues that Buzuku’s admirative could be interpreted as having a present meaning, but, aside from the fact that both Latin (servasti) and Greek (τετήρηκας) as well as Bible 1872 and Bible 1930 all have perfects, the most telling evidence is Bible 1980. The translator of Bible 1980, Simon Filipaj, had access to Buzuku, and it would appear that here he was influenced by Buzuku’s language in choosing an admirative. If the “present” admirative had already shifted to a true present meaning, we would expect Filipaj to have preserved it. The fact that he uses a perfect admirative indicates that Buzuku’s ruojtkē was still a marked perfect, albeit one that shows evidence of the modern admirative meaning of surprise from direct observation.
In the one indicative imperfect admirative, Mary Magdalene goes to Jesus’s tomb and finds the stone removed (and the body gone) and goes running to the disciples and says (again, see above regarding (6.185) and Footnote footnote 311):
a.
Marrëkishnë Tenëzonë ṇ vorit … (Buzuku LXVI/2 = Çabej 1968) take.inv.plu.3pl b.
Ngrĭtnẹ Zŏtinẹ prei vŏṙit … (Bible 1872) moved.3pl c.
E ngritnë Zotinë nga varri, edhe nuk dimë se ku a kanë vënë (Bible 1930) moved.3pl d.
E kanë marrë Zotrinë prej varrit … (Bible 1980) have.3pl take.ptcp ‘They have taken away the Lord out of the tomb [and we don’t know where they (have) laid him]’ (John 20:2)
Here the Latin has a perfect (tulerunt) and the Greek an aorist (ἦραν). Bible 1872 and Bible 1930 follow the Greek and Bible 1980 follows the Latin. The sense of surprise at a discovery is certainly present, but an imperfect is clearly not called for. Demiraj interprets it as an expressive, inverted pluperfect. Here we can also say that the use of an inverted pluperfect with perfect meaning parallels the eventual use of the inverted perfect with true present meaning. What can be seen here, then, is a glimpse of the semantic processes that led to the eventual development of the Albanian admirative as we know it today. This type of usage survives in the fact that if the utterance expresses surprise at a state of affairs that existed prior to the speaker’s discovering it, modern Albanian permits any of the four admirative paradigms. It is precisely and only in such contexts that the Albanian admirative corresponds to admirative usage in Bulgarian, Macedonian, and Turkish (see Friedman 1981, 2005a for discussion), a problem to which we return below.
The final example here is the most problematic of all, and also gives possible insight into Buzuku’s sources. Here, Mary Magdalene anoints Jesus’s feet with expensive ointment, and Judas asks why the ointment is not sold and the money given to the poor, and Jesus replies (once again, (6.185) and Footnote footnote 311 are relevant here):
a.
Lee atë ṇ ditet së vorit tim ajo kët paskā ruojtunë. (Buzuku LVI = Çabej 1968) have.adm.3sg keep.ptcp b.
Ľen-e; se e rŭeiti pẹr ditẹn’ e sẹ voṙŭemesẹ s’eme. (Bible 1872) keep.aor.3sg c.
Lër-e; se e ruajti për ditën’ e të-kallurit-tim. (Bible 1930) keep.aor.3sg d.
Lëre! Le ta bëjë këtë për ditën e varrimit tim! (Bible 1980) let dms.it do.sbjv.3sg ‘Leave her alone! Let her keep it/She has kept it for the day of my burial.’ (John 12:7)
Here, as Çabej 1968: I.37–38 correctly notes, the Latin (Vulgate) has a subjunctive ([ut …] servet) but the Serbian and Italian versions both have past (je … dohranila and ha serbato, respectively). Çabej also cites the Greek ἐφύλαξεν, an aorist that is used in the Greek Orthodox version. The Received version has a perfect (τετήρηκεν) whereas revised editions follow variant readings with the aorist subjunctive (τηρήσῃ).Footnote 360 Given that the overwhelming majority of admiratives in Buzuku are modal, it could be that the meaning here is subjunctive rather than perfect, although the lack of a modal marker argues against this interpretation. It might be, then, that this is a true perfect admirative, but even if it is, its isolated nature indicates that the shift from ontological counterfactual to personal counterfactual, i.e., nonconfirmative (since surprise implies the existence of a counterfactual expectation, sarcasm is a genuine personal counterfactual, and both inference and report allow for the possibility that the statement might not be factual), has only just begun.
Diagnostic of the indicative inverted perfects transfer from past to present admirative is the usage found in the Albanian dialects of Ukraine, whose speakers separated from the main body of Tosk in the seventeenth century, migrated eastward to the Black Sea region of Bulgaria, and then northward through what is today Romania and Moldova and ended up in Ukraine. They separated from the main body of Tosk precisely when the inverted perfect was still perfect but already had the nuances of the admirative complex. Example (6.196) shows that what is today a present admirative in Albanian still has the value of a nonconfirmative perfect in the dialect of Ukraine.
(6.196)
Ato rojtkan ni Albanijet. Turku i mëndojka shumë malserivet, çupa dhe gra i përçumojtka. Ato gandej ikkën. Vatkan ni Serbijet. Sa rojtkan ni Serbiet nok a di. Ga Serbija rakan ni Bullgariet. Ato ni Bullgariet rojtkan ni Devnat. Juni fshat thuhet Devna. Ga Bullgarija ato dalkan dhe rakan ni Molldavien. Nok a di çilittë motmot. Rossija a mur Molldavien. Dhe ga Molldavija ni mila tetë kindë gjashtëdhita na jerme ni Tavriçeskaja gubernija. (Voronina et al. 1996: 179) ‘They lived (adm) in Albania. The Turk oppressed (adm) the mountain folk a lot, he dishonored (adm) girls and women. They left (aor) from there They went (adm) to Serbia. How long they lived (adm) I don’t know. From Serbia they came (adm) to Bulgaria. In Bulgaria they lived (adm) in Devna. Our village is called (prs) Devna. From Bulgaria they went out (adm) and came (adm) to Moldavia. I don’t know (prs) what year. Russia took (aor) Moldavia. And from Moldavia in 1860 we arrived (aor) in Tavricheskaja Province.’
It is worth noting that among the Eastern Diaspora Albanian dialects of Greek Thrace documented by Liosis 2021, some have admiratives that look like standard Albanian, while others have lost these constructions altogether.
To sum up: the Albanian admirative is a morphosyntactic category whose use straddles the Geg/Tosk divide (cf. Mindak 1986). The development of the inverted perfect into a classic admirative set of paradigms (in which the inverted perfect itself becomes a true present and a new admirative perfect is built using the present admirative auxiliary with a participle, with the same relationship of imperfect to pluperfect in the fullest paradigm) is especially characteristic of central Albania (from Central Geg to Northern Tosk), where urban centers such as Elbasan were dominated by Turkish. Here, as elsewhere in the Ottoman Balkans, Turkish functioned as a marker of urban identity, and being a town resident meant being able to speak Turkish (see Ellis 2003). In Northwest Geg, the admirative still retains nuances of its meaning as an inverted perfect, especially in rural areas (Çabej 1979: 16–18), and it even shows a tendency toward being eliminated via restrictions on its occurrence. Thus, for example, in Dushmani, 30 km east of Shkodër, near the Montenegrin border, the admirative only occurs in the perfect, e.g., pàska pà[s], Standard Albanian paska pasur (Cimochowski 1951: 116). In East Central Geg, the present and perfect admirative are viable, but the imperfect and pluperfect are either rare or absent (Hoxha 1975: 167, Hoxha 1990: 139; Murati 1989: 43; Bashi 1989: 192). The admirative is absent from the Lab and Çam dialects of the extreme south of Albania and adjacent parts of northern Greece (Altimari 1994, cf. however Totoni 1971: 74). Also suggestive is the fact that the admirative is absent from the Arbëresh dialects of Italy, and the Arvanitika dialects of Greece, which separated from the main body of Albanian before contact with Turkish, have retained only traces of the modal usage described above (Liosis 2010). On the other hand, as seen in (6.196), in the Albanian dialects of Ukraine, which separated from the main body of Albanian in the seventeenth century, after approximately two centuries of Turkish contact, there is an inverted perfect, but it is an evidential past and not a present admirative (Kotova 1956). It can thus be argued that the meanings of nonconfirmativity (i.e., surprise, doubt, report, inference) associated with modern admirative usage developed during the Ottoman period, radiating from the center outward in both Geg and Tosk. In this scenario, the encoding of evidentiality represented by the admirative, while built on native material, probably did not enter the grammatical system until after contact with Turkish. For Albanian (as for Balkan Slavic, whose perfects acquire nuances similar to those found in the admirative), it can be hypothesized that the grammatical encoding of evidentiality began in urban centers where Turkish was widely spoken and had high prestige and spread from there to the countryside. It is worth noting that the admirative is particularly viable in Kosova, where the prestige of Turkish in the towns lasted well into the twentieth century. See also §6.2.4.3.1.2 on the use of qenka instead of qoftë in the Albanian of North Macedonia on analogy with Macedonian bilo.
Thus, it was precisely during the Ottoman period, and precisely in the dialects that separated from the main body of Albanian a century or two after the Ottoman invasion (i.e., centuries after Arbëresh migrations) that the stage of inverted perfect as nonconfirmative equivalent to the Turkish mIş-past is preserved.
6.2.5.6 Aromanian: The Frasheriote Dialect of Bela di Suprã
The Frasheriote Aromanian dialect of Gorna Belica (Aro Bela di Suprã/Bela di Sus) has a specific set of paradigms not yet attested elsewhere, although equivalent, but as yet undescribed, dialects might still exist in Albania.Footnote 361 In this Frasheriote dialect, speakers have calqued the Albanian aorist of ‘be’ as an auxiliary, they have calqued the compound analytic tenses, and they have reanalyzed the Albanian 3sg.prs admirative marker -ka as an invariant marker of nonconfirmativity and attached it to a m.pl.impf participial base. Three or four verbs use the aorist base, most importantly ‘be’ and ‘have,’ but this can be attributed to the identifiably aorist stem of the participle in these common verbs in both Albanian and Aromanian. In the Albanian source dialects (Myzeqe), the past participle ends in -ë, which looks like a masculine plural suffix, whence, presumably, the otherwise unused m.pl.impf participle as the base. Table 6.31 gives representatives of the relevant admirative paradigms in the two languages as well as the indicative calqued forms. Albanian has the possibility of using both the aorist and the imperfect of ‘have’ and ‘be’ to form pluperfects, and, moreover, it can use analytic perfects and pluperfect as auxiliaries. This is followed by illustrative examples, and then a brief discussion of the participial bases.
| Nonadmirative | Admirative | |||
|---|---|---|---|---|
| Aromanian | Albanian | Aromanian | Albanian | |
| Present | lukrã | punon | lukracka | punuaka |
| Perfect | ari lukratã | ka punuar | avuska luktratã | paska punuar |
| Imperfect–pluperfect | ave lukratã | kish punuar | – | paskësh punuar |
| Aorist–pluperfect | avu lukratã | pat punuar | – | – |
| Perfect–pluperfect | ari avut lukratã | ka pasë punuar | ari avuska lukratã | paska pasë punuar |
| Imperfect–pluperfect 2 | ave avut lukratã | kish pasë punuar | ave avuska lukrată | paskësh pasë punuar |
| Aorist–pluperfect 2 | avu avut lukratã | pat pasë punuar | – | – |
Example (6.197) gives the original Macedonian text in (b) that yielded the Aromanian translation (a) that resulted in the discovery of this Aromanian admirative (see Friedman 1994b):
a.
Un a meu oaspit bănædză tu Bitul’i, ma di multu a to me friend lives in B. but from much oară nu n(ə) avem vădzută. Aseră vini năs(nəs) la times not us we.have seen.f yesterday came he to mini ku Mercedes. Mi l’ oi̯ di minti [Mi me with Mercedes me it took from mind [me čudosi]: Abe tora tini fuska avut om! (Aro) amazed] voc now you are.adm rich man b.
Eden moj prijatel živee vo Bitola, no odamna one my friend lives in B. but long.time.ago se nemame videno. Včera toj dojde kaj mene so intr not.we.have seen.n yesterday he came by me with Mercedes. Se šašardisav: Abe sega ti si bil bogat čovek! (Mac) Mercedes intr I.wondered voc now you are was.n rich person ‘A friend of mine lives in Bitola, but we had not seen one another for a long time. Yesterday he came to my place in a Mercedes. I was amazed: Hey, you are a rich man now!’ (August 1992)
The following additional examples (some with their Macedonian equivalents, and for (6.204), the corresponding Albanian) are reported by Markovikj 2007: 143 (and also collected via fieldwork), once the form had been discovered.Footnote 362
a.
Kum di nu, muntretskã televizije nalili (Aro) how of neg watches.adm television news b.
Kako da ne, toj gledal televizijski vesti (Mac) how dms neg he watch.lf.m television news ‘Oh sure, he’s watching/watches TV news’ (not!)
a.
Kum di nu, vãzê vãnetska /u s-vãneckã (Aro) how of neg fut come.adm / FUT dms-come.adm ‘Oh sure, (he said) he’ll come’ (not!) b.
kje dojdel / kje došol (Mac) fut come.lpt / fut come.lpt ‘Oh sure, (he said) he’ll come’ (not!)
(6.200)
Kum di nu, u štietska Viktor pi Kiro Gligorov (Aro) how of neg intr know.adm V. on K. G. ‘Oh sure, Victor knows Kiro Gligorov (and pigs can fly).’
a.
Abe Toma avuskã dzãsã s-nu kãntãm kãntuc de rãmãneshti (Aro) voc Toma has.adm said dms-neg sing.1pl songs of Aromanian b.
Abe Toma imal rečeno da ne peeme vlaški pesni!? (Mac) voc Toma has.lf said dms neg sing.1pl Aromanian songs ‘Huh?, Toma has said we shouldn’t sing Aromanian songs!?’
a.
Tu komshats avea unã kasã, tsi irea at neighbors have.impf a house that were.impf multu oarfãn (Gołąb 1984a: 184; Mbãliot) very poor b.
Tu kumshăts av’ unã kasã, tsi fuska multã orfãnã. (Frasheriote) at neighbors have.impf a house that be.adm very poor c.
Komšiite imale kukja, što bea/bila (Mac) neighbors.def.pl had.lf.pl house that were.impf/was.lf.f mnogu siromašni very poor ‘The neighbors had a house that … they were/it was very poor’
(6.203)
tini [ai / avuska ~ avuska avutã] mult oi (Aro) you have have.adm have.adm had many sheep ‘You have [statement of fact / surprise, report, etc.] a lot of sheep’
a.
Abe, munduem/*munduesk ka Silja kãntatska! (Aro) b.
Abe, mislev/*mislam deka Silja peel (Mac) c.
Ore, mendova/*mendoj se Silja kënduaka! (Alb) voc I.thought/*I.think that Silja is singing ‘Hey, I thought Silja was singing/was supposed to sing’ (on seeing someone else)
Table 6.32 illustrates the arguments for positing the way in which the Albanian admirative participial base was interpreted in Gorna Belica Frasheriote Aromanian (GBFA). The GBFA admiratives of ‘be’ and ‘have’ have the variants fu-ts/s-k[ã]r]ã ‘be’; avuska, avuskãra, avuskra ‘have,’ to which can be added the GBFA admirative of ‘see’ vãdzuska, vãdzucka. These three verbs do not use the imperfect (or present) stem: *areska, *aveska, *vadeska; and these three verbs show suppletion in the aorist in Albanian.
| gloss | 1sg.prs | 1sg.impf | 1sg.aor | ptcp | adm |
|---|---|---|---|---|---|
| have | kam | kisha | pata | pasur | paska |
| be | jam | isha | qeshë | qenë | qenka |
| see | shoh | shohja | pashë | parë | paka |
All other GBFA verbs use the imperfect stem, specifically the masculine plural imperfect past participle, for example:
(6.205) nãdzeska (*neska) ‘go, walk’
dusetska (*duska) ‘go, lead’
vanetska (*vãnicka) ‘come’
kunushteska (*kãnãskuska) ‘know’
betska (*bitska) ‘drink’
In Tosk Albanian, the schwa (orthographic <ë>) that occurs on many (past) participles is homophonous with a masculine plural adjectival ending. Also, a number of variants occur in dialectal Tosk for the admirative, as seen in material from various sources:
a. Pekmezi 1908: 198: hǘpkërkam < hǘpërkam < hǘpkam ‘get up’; ubëkërkam < ubë’rkam < ubë’kam ‘become’
b. Lambertz 1948: qënkërkam ‘I am’ (qenë), paskërke, pasërke ‘you have’ (pasur), dhënkërke ‘you give’ (dhënë), you ditkërke ‘know’ (ditur), kuptokërke ‘you understand’ (kuptuar), dashërka ‘want’ (dashur), vatërka ‘s/he goes’ (vajtur) ‘go to’
Thus a GBFA form, e.g., of ‘have,’ would be avutsi + ka just like Albanian pasurë + ka; basically, then, the Tosk variants account for the GBFA variants.
We can also mention here (see also §6.2.5.1) that modern Boboshtica (Mac Boboščica) Macedonian, which has been in contact with Albanian for many generations – as was Frasheriote Aromanian – currently has an interesting development influenced by the Albanian admirative (the data in this paragraph are based on Makartsev 2013ab).Footnote 363 When Makartsev asked a consultant to translate Albanian paskam marrë paratë ‘I have taken (adm) the money,’ which Makartsev translates into Russian as èto ja-to vzjal den’gi (‘Yeah, sure, me, I took the money!’), and then to conjugate the verb ‘take’ in the form he produced, the following paradigm emerged: The singular is invariant imal vziäto ‘I, you[sg], s/he, it have taken’ but the plural shows especially interesting variation. The first two persons have variants, 1pl imalme vziäto/imalo vziäto and 2pl imalte vziäto/imalo vziäto, while the 3rd pl has only imalo vziäto. Here we have a reinterpretation of the l-form as a participial base, on the Albanian model, although the l-form had already ceased to be participial in the sense discussed above (cf. Table 6.28). The present person markers in the 1pl/2pl are reinterpreted as mobile morphemes that can attach to the l-form, the 1st/2nd person auxiliaries having gone the way of the third person auxiliaries throughout western Macedonian. While distinct from the Bela di Suprã data cited here, it is related in being influenced by Albanian to create a new paradigm. (See also §6.2.5.2.)
6.2.5.7 The Romanian Presumptive
The so-called presumptive mood of Romanian has the same complex of nonconfirmative meanings discussed above (admirative/dubitative/report). It can be formed with any of the three paradigm-forming modal markers – future o (as well as conjugated va and variants), subjunctive să, and conditional aş, etc.) – plus invariant fi ‘be’ with either the gerund or past participle, as seen in Table 6.33, for the verb a lucra ‘to work.’
| Future | voi-vei-va-vom-veți-vor ~ o ~ i ~ oi-oi-o-om-oți-or | fi | lucrând | present |
| Subjunctive | să | lucrat | past/perfect | |
| Conditional | aş-ai-ar-am-ați-ar |
The following examples illustrate uses of the Romanian presumptive and how they relate both to other Balkan languages and to the rest of the Romanian verbal system. Examples (6.207)–(6.210) all employ the gerund and are unambiguously presumptive, as there is no other possible interpretation (cf. Popescu 2017: 225). Moreover, these examples show that all three modal particles (dms, fut, and cond) can be deployed without any significant semantic differentiation.Footnote 364
(6.207)
Mihai, oare să fi existînd strigoi? M. really dms be existing ghosts ‘Mihai, do ghosts really exist?’ (Vasiliu 1966: 224)
(6.208)
– Îți zice lumea “Niculăiță Minciună? – Mi-o fi zicînd – you.acc calls.3sg the.world N. the.liar – me-fut be calling – Do they call you Nick the liar? – They supposedly call me that.’(Vasiliu 1966: 224)
(6.209)
Doar, n-or fi avînd purici! surely neg-fut be having fleas ‘Surely he doesn’t have fleas!’ (Ioanna Chitoran, p.c., February 1998)
(6.210)
Pe lîngă urs se spune că ar fi avînd și această on along bear intr says that has.cond.3sg be having and this pajură care-l priveghează. eagle that-him guards ‘Alongside the bear it is said that there is a golden eagle that keeps a vigil over him.’ (Vasiliu 1966: 224)
Examples (6.211)–(6.219) address the fact that with the (past) participle (as opposed to the (present) gerund), the presumptive is homonymous with the anterior future (with va, etc.), past subjunctive (with să), or past conditional (with aş, etc.). Examples (6.211)–(6.214) are past presumptive, (6.215) is a contrasting anterior future, (6.216) is a presumptive showing that all three modal particles are, in this context, equivalent, (6.217) is an anterior future, (6.218) contains both a past (counterfactual) conditional and a similarly past subjunctive, and (6.219) is past (counterfactual) conditional in both the protasis and apodosis.Footnote 365
(6.211)
Iar doamna lui Dragoş-vodă, aşa povestescu oamenii and lady.def of.him D.-king thus say.prs.3sg people acei de locu, de la târgul Siretului, those of place.def of in market.town Siret.gen.def cum să fie fost de leage sască how dms be been of religion Saxon ‘and king Dragoş’s lady, as the people of the place, the market town of the river Siret, say, was of Saxon religion.’ (Moldavian chronicle, Manoliu-Manea 1994: 310)
(6.212)
Doar la Iaşi să fi fost aşa ceva only in Iaşi dms be been such something ‘Only in Iaşi could such a thing have been’
(6.213)
Zice că ar fi citit lecția he.says that have.cond.3sg be read the.lesson ‘He says that he has read the lesson’ (admirative/dubitative/reported)
(6.214)
Va fi citit el acest roman? [Mă îndoiesc.] fut.3sg be read he this novel me.acc doubt ‘Has he read this novel?/(He claims he read this novel.) [I doubt it.]
(6.215)
Vom sta de vorbă numai după ce-l vei fi citit şi tu fut.1pl stand of speak only after that-it fut.2sg be read and you ‘We’ll have a chat (about the novel) only after you have read it, too’
(6.216)
va / să / ar fi ajuns el pîna acolo? fut.3sg / dms cond.3sg be arrive he until there ‘Did he arrive here [presumptive]?’
(6.217)
cînd ea va fi acasă, el va fi ajuns pîna acolo when she fut.3sg be at.home he fut.3sg be arrived up.to there ‘When she gets home he will have arrived there.’
(6.218)
în alte condiții nu ar fi fost posibil să fi ajuns in other conditions neg cond.3sg be been possible dms be arrived el pîna acolo he up.to there. ‘Otherwise it wouldn’t have been possible for him to have arrived there.’
(6.219)
dacă ar fi ajuns el pîna acolo, alta ar fi fost situația if cond.3sg be arrived he up.to there other cond.3sg be been the.situation ‘If he had arrived there things would have been different.’
Thus, while the past presumptive can be compared to modal usages (especially of conditionals) in other Romance languages, the present presumptive behaves very much like – albeit not identically to – so-called evidential forms in Balkan Slavic, Albanian, and Turkish (which also differ among themselves, as seen above). Examples (6.220) and (6.221), which consist of Romanian and Turkish translations from the Bulgarian novel Bai Ganyo, illustrate particularly clear correspondences between the Romanian modul prezumtiv, on the one hand, and the Bulgarian and Turkish evidentials, on the other. The relevant forms are bolded and underlined in each language, and minor differences in the translation have been ignored.Footnote 366
a.
Baj Ganjo poiska az da počerpja, poneže iz pătja săm pušil ot negovija tjutjun. (Konstantinov 1895: 81) (Blg) b.
Bai Ganiu îmi ceru săi fac cinste, deoarece pe drum aş fi fumat din tutunul lui. (Konstantinov 1964: 96) (Rmn) c.
Büfede, Bay Ganü, yenip içilecek şeyleri benim ısmarlamam gerektiğini ileri sürdü. Çünkü yolda hep onun tütününden içmişiz. (Konstantinov 1972: 128) (Trk) ‘Bai Ganyo wanted me to treat, since on the way I had supposedly smoked some of his tobacco.’ (Konstantinov 2010)
a.
Kato dojde tazi godina s gluhonjamoto momče, kaza mi, če tuj momče imalo u vas, v Bălgarija, brat činovnik ili oficerin – ne pomnja, – kojto štjal da mu otpušti po sto franka ežemesečno … (Konstantinov 1895: 82) (Blg) b.
Cînd se întoarse, veni cu surdomutul, îmi spuse cum că băiatul ar fi avînd în Bulgaria un frate ofițer, ori funcționar – nu mi-a duc aminte –, care avea să-i trimită cîte o sută de franci pe lună … (Konstantinov 1964: 97) (Rmn) c.
Bu yıl, beraberinde sağır ve dilsiz çocukla geldiği zaman bana şunları yutturdu: Sözde bu çocuğun Bulgaristan’da memur mu, subay mı, pek iyi hatırlayamıyorum, ağabeysi varmış. kardeşine her ay yüz frank gönderecekmiş. (Konstantinov 1972: 129) (Trk) ‘When he showed up this year with the deaf-mute, he told me that the kid had a brother in your country, in Bulgaria, an official or an army officer – I don’t remember – who was supposedly going to send him a 100 francs a month … ’ (Konstantinov 2010)
In example (6.220), the Bulgarian original and Turkish translation make it clear that the meaning is nonconfirmative; in Romanian, the form is cond + fi + ptcp. In example (6.221), Bulgarian imalo (reported) corresponds to a present conditional presumptive, whereas the reported future (štjal da otpušti) corresponds to a ‘have’ anterior future in the Romanian translation. Turkish uses -mIş in both instances.
Finally, (6.222) illustrates a Romanian past presumptive (again with the conditional marker) that corresponds to a nonconfirmative in Turkish, Balkan Slavic, and Albanian, albeit in different ways. Romanian cică, Bulgarian kaj, Macedonian veli, and Albanian gjoja, are all reportative or dubitative particles. In the Albanian, it is the ‘it isn’t possible’ part that is expressed in the nonconfirmative (admirative), whereas in all the other languages there is a verb of reporting that is placed in the nonconfirmative to render the same effect.
(6.222)
Cică i-ar fi spus … “Nu se poate … ” (Konstantinov 1964) (Rmn) Ne može, kaj, kazal (Konstantinov 1895) (Blg) Ne može, veli, mu rekol (Konstantinov 1967) (Mac) Nuk qënka e mundur, gjoja (Konstantinov 1975) (Alb) “ … olunmaz,” demiş (Konstantinov 1972) (Trk) ‘“It’s not possible” he1 says he2 said’ (Konstantinov 2010)
This example demonstrates that while each of these Balkan languages deploys a combination of lexical and grammatical means to express nonconfirmativity, the actual manifestations are language specific. It is precisely such examples that demonstrate the need for a nuanced understanding of the commonalities and differences among the Balkan languages.
6.2.5.8 The Aromanian Presumptive
We can also mention here an Aromanian presumptive, which has a variety of definitions. For us the crucial syntagm is fut ±dms + aor. However, since the Aromanian equivalent of the Romanian presumptive has received attention only in the Aromanian dialectological literature, we give it special attention here. The main point is that fut ± dms + aor is specifically western Balkan (Aromanian, Macedonian, Greek, Albanian) and not eastern Balkan (Bulgarian, Romanian), although other syntagms are also included under this label in various descriptions. Nonetheless, we give a complete overview in this section.
6.2.5.8.1 The Aromanian Presumptive: Description
The earliest discussion of the Aromanian presumptive appears to be Capidan 1932: 470, followed by Caragiu-Marioțeanu 1968: 110–111; Saramandu 1984: 459; and Nevaci 2006: 153.Footnote 367 The exposition below refers to these authors followed by evidence from Markovikj 2007: 101; Gołąb 1984a:107; and Bara et al. 2005: 194, 210–212, and concludes with the Balkanological implications of the data.
Capidan, who labels the paradigms viitorul anterior ‘anterior future,’ only gives the paradigm for the future perfect (va s- + subjunctive perfect), but in the notes he includes examples with va(i) ± s- plus imperfect, aorist, and pluperfect as well. The examples with perfects and aorists are translated into Romanian using the anterior future presumptive (va fi + [past] participle) while those with the imperfect and pluperfect, which Capidan calls a past optative, are translated using the past conditional presumptive (aş [etc.] fi + [past] participle). Capidan also notes that s- can be omitted after va, especially when followed by a pronoun, but, as his examples illustrate, sometimes va (or vai) can also immediately precede the verb. Of interest to us here, as is seen below, are the aorists, as in the following examples from Papahagi 1905, the verb forms of which are cited by Capidan (his Romanian translations for the verbs are given in square brackets):
(6.223)
Vîră picuràr vaĭ aprease focul [va fi aprins] (Papahagi 1905: 386/6) some shepherd fut light.3sg.aor fire.def fut be light.pst.ptcp ‘Some shepherd must have lit the fire’
(6.224)
minduì c-aestă va-l vătămă´ think.3sg.aor that-this.f fut-it kill.3sg.aor fičorlu [îl va fi omorît] (Papahagi 1905: 283/35) child.def it.acc fut.3sg be kill.pst.ptcp] ‘he thought she (lit., ‘this one’) would kill the child’
(6.225)
va-ńĭ adràşĭ groapă! [îmi vei fi făcut] (Papahagi 1905: 469/27) fut-me.acc do/put.2sg.aor hole [me.acc fut.2sg be make.pst.ptcp] ‘you’ve put me in a hole!’ [said by a devil suddenly falling into a trap]
(6.226)
Ma tse va s-featse mul´earea dzîse lamnja but what fut dms-do.3sg.aor woman.def say.3sg.aor monster.def [va fi făcut] (Papahagi 1905: 407/28) [fut.3sg be do.pst.ptcp] ‘But what has the woman done? said the monster.’
Examples (6.223) and (6.224) illustrate conjectures, while (6.225) and (6.226) involve admirative usages. All the examples are from what is now Greece (Epirus or Aegean Macedonia).
Caragiu-Marioțeanu, who was born in Hrupishte (Grk Árgos Orestikon, Aegean Macedonia), uses the label viitorul II ‘future II’ and gives complete paradigms for all four tenses with glosses in Romanian (our English glosses are approximate, since context would determine the exact gloss and no contexts are given), e.g., perfect va s-amu cîntată ‘voi fi cîntat’ ‘I must’ve/would’ve sung,’ aorist va-lu vîtîmaĭ ‘l-oi fi omorît, îl voi fi omorît’ ‘he must’ve/would(‘ve) killed him,’ imperfect va vîtîmám ‘era să-l omor, l-aş fi omorit, eram în stare să-l omor’ ‘apparently I was going/able to kill him,’ and pluperfect va avḙámu vîtîmată ‘I had killed him,’ which she says has the same value as the pluperfect. As can be seen, all tenses include va, but s- is included only for the perfect and is absent in the other paradigms. She also states that only with the perfect is the paradigm a future in the past and adds that va + aorist is a past presumptive, while va + imperfect or pluperfect are past conditionals expressing probability or possibility.
Saramandu, who collected his examples in the field (apparently in Greece), uses the label dubitativ (presumptiv) ‘dubitative (presumptive)’ and distinguishes a present, perfect, and pluperfect. The present is s-+aorist, which is only interrogative, as seen in (6.227). The perfect is the present dubitative preceded by va, as in (6.228), and the pluperfect is formed with va s- + subjunctive perfect, as in (6.229) (all with Saramandu’s Romanian translations, in brackets, all from Saramandu 1984: 459):
(6.227)
si-lu feĉu mini? [să-l fi făcut eu?] dms-it.acc do.1sg.aor I dms-it.acc be done.pst.ptcp I ‘Did I do that?!’
(6.228)
va s-cîntáĭ [oi fi cîntat] fut dms-sing.1sg.aor fut.1sg be sung.pst.ptcp ‘Apparently I sang.’
(6.229)
va s-amu cîntatî [oi fi cîntat (lit., *oi fi fost cîntat)] fut dms-have.1sg.prs sing.pst.ptcp fut.1sg be sung fut.1sg be been sung ‘Apparently I sang/had sung.’
Nevaci follows Saramandu’s description, adding that the dubitative meaning is superposed on that of the anterior future, subjunctive perfect, and subjunctive-optative perfect.
As indicated above, the construction va(i) ± s- + aorist seems to be limited to southern Aromanian (Greece). Markovikj 2007: 101 has no examples of u/va + s[ă] + aorist from the Ohrid-Struga region. He notes that the future marker can occur with the present, imperfect, pluperfect, and perfect, and he treats these as more or less equivalent to the same constructions in Macedonian. Gołąb 1984a: 107 writes explicitly that the aorist does not occur with the subjunctive marker, and Capidan’s 1932: 470 sole example from Crushova (Mac Kruševo) uses a perfect, which is typical for any Balkan language.
Bara et al. 2005: 194, 210–212 record instances of s’ + aorist for the Aromanian dialect of Turia (Grk Krania), for which they provide Greek translations. These usages are all negated: (6.230) is interrogative with dubitative nuance, (6.231) is exclamatory-interrogative dubitative, and (6.232) involves the protasis of an irreal conditional, as can be seen from the corresponding Greek and the glosses. Examples (6.230) and (6.231) resemble (6.227), but (6.232) is quite unusual, and its ungrammaticality hinges on the use of the aorist έπεσε:
(6.230)
ś-nu murì màrkulu? Να μην πέθανε ο Μάρκος; (Bara et al. 2005: 194) dms mneg died M.DEF ‘Marko didn’t die, did he?’
(6.231)
ĭo ş nu-ľ aģutài? Εγώ να μην τον βοήθησα; (Bara et al. 2005: 194, 211) I dms mneg him helped.aor ‘What, as if I didn’t help him?’
(6.232)
va nă ćeàmu la mitèori, ka ś-nu kădeà neàuă. Θα πηγαίναμε στα Μετέωρα, να μην *έπεσε το χιόνι. (Bara et al. 2005: 194) fut went.1pl to.the M. dms mneg fell.aor the snow ‘We would have gone to Meteora if it hadn’t snowed.’
6.2.5.8.2 The Aromanian Presumptive vis-à-vis the Romanian
As can be seen from the data presented in §6.2.5.8.1 the label presumptive covers a complex set of phenomena in Aromanian. Here it is worth noting that the Romanian prezumtiv differs from its Aromanian equivalent in three important respects. First, in Romanian the main verb is nonfinite and can be either present (gerund) or past (participle) and is accompanied by the nonfinite auxiliary fi, whereas in Aromanian the main verb is finite (this includes the analytic paradigms involving conjugated ‘have’ plus invariant participle). Second, the Romanian presumptive involves either the subjunctive marker să, the future marker o ~ voi, etc., or the conditional marker aş, etc., whereas the Aromanian involves only the future and/or subjunctive markers. Third, the Romanian present presumptive (with the gerund) is unambiguously presumptive, while the Romanian past presumptive (with the participle) is ambiguous between presumptive meaning and past subjunctive, anterior future, or past conditional, depending on the marker (see Friedman 1988c, 1997b on disambiguation and also for the argument that all three sets of markers do indeed involve presumptivity). In contrast to Romanian, Aromanian has no present presumptive per se, but the Aromanian aorist can occur after future and subjunctive markers, a syntagm that does not occur in Romanian. It is the Aromanian use of the aorist after subjunctive and future markers that constitutes its distinctive presumptive. As in Romanian, future or subjunctive markers with imperfect, perfect, or pluperfect have the potential for other modal meanings. This fact is explicitly recognized by Caragiu-Marioțeanu. Thus, in what follows, we turn our attention specifically to subjunctive and future markers with the aorist in Aromanian.
A question raised by the accounts cited in §6.2.5.8.1 is the distinction between s- + aorist, and va ± s- + aorist. Saramandu contrasts s- + aorist with va s- + aorist, while Bara et al. 2005 only have examples of s- + aorist, Capidan has va ± s- + aorist, and Caragiu-Marioțeanu gives only va + aorist in the main body of her text, although she cites Capidan’s forms in a footnote. Judging from the available data, we can say that in general s- + aorist is limited, for the most part, to dubitative-admirative interrogative exclamations, i.e., they are affective expressions denoting either disbelief or surprise as seen in (6.227), (6.230), and (6.231). This type of usage is also seen with va ± s- in examples (6.225) and (6.226) as well as (6.227). Examples (6.223), (6.224), and (6.228), however, show a nonexpressive type of presumptive meaning. Example (6.232) is unusual in that it is the protasis of a conditional period, a meaning that is neither expressive nor presumptive.
6.2.5.8.3 The Aromanian Presumptive as a Balkanism
The Aromanian presumptive as defined here consists of an aorist subordinated to the subjunctive marker (s-) or the future marker (va ± s-). While other past paradigms can also occur after these markers, the others are all ambiguous, as are the corresponding forms of the Romanian past presumptive. The Aromanian presumptive based on the aorist seems to be limited to southern Aromanian, where Greek is the main contact language. Moreover, like Aromanian, Greek permits aorists after the subjunctive marker να, as well as its negated form (να μην), which is seen in (6.230) and (6.231), as well as after the future marker (θα), and, moreover, these constructions are likewise expressive. Albanian also permits the aorist after markers of future (do (të)) and subjunctive (të, usually preceded by po), but such usage generally occurs in conditional periods, not unlike example (6.232). In Romani, since there is no native perfect/aorist distinction (see Table 6.16), te + nonremote perfective past (in Matras’s 2002 terms) is unremarkable. Macedonian permits the aorist only after the expression kako da ‘as if,’ but this differentiates it from Bulgarian, which never permits the aorist after da (for ‘as if,’ Bulgarian has kato če). Thus, we can say that the structure of the Aromanian presumptive adds to Gołąb’s 1976 distinction between eastern and western Balkan Slavic and Balkan Romance, where Bulgarian and Romanian, on the one hand, and Macedonian and Aromanian, on the other, share various convergences. Here the innovation in Aromanian and Macedonian involving a type of modal aorist is shared – albeit with different sets of restrictions in each language – with Albanian and Greek. Thus, as with many other linguistic features, the modal aorist can be seen as a heartland Balkanism (cf. Hamp 1977a), i.e., one that arose in the context of the contacts among speakers of various languages in the general area of the Vilayet of Manastir and neighboring sandzhaks and that achieved its current form during the Ottoman period.
These forms suggest that the developments that led to the Romanian presumptive were in fact a post-Ottoman phenomenon, independent of these Aromanian forms, which, in any case, are not well attested.
Finally, we can note that in addition to the ordinary future marked by the dms or ãs (see §6.2.4.1), Meglenoromanian also has a construction that is analyzed as an equivalent of the Romanian presumptive, which involves the full ‘want’ auxiliary, va, plus infinitive (Atanasov 2002: 230–238, 248). An example such as (6.233a) from Atanasov 2002: 230, given here with its Romanian translation (b), appears to fit this description:
a.
va veári múltu lúcru di nu poáti viníri (Megl) fut have.inf much work of neg can.3sg.prs come.inf b.
o fi având multu de lucru de nu poate veni (Rmn) fut be.inf have.grd much of work of neg can.3sg.prs come.inf ‘Apparently he has a lot of work so he can’t come’
However, va + inf in Aromanian can also be a suppositional that does not involve a presumptive in Romanian, and is closer to the kje da construction in Macedonian as in (6.234a) from Atanasov 2002: 230, given here with its Romanian translation (b):
a.
va vḙári dói ań di căn nu nă vem vizútă (Aro) fut have.inf two years from when neg we.acc have.3sg.prs see.ptcp b.
or fi doi ani de când nu ne-am văzut (Rmn) fut be.inf two years from when neg we.acc-have.1pl.prs see.ptcp ‘It will be (= must be, cf. Mac kje da ima) two years since we’ve seen each other’
6.2.5.9 Novo Selo Bulgarian
The Bulgarian dialect of the village of Novo Selo, Vidin region, has a special mood called the probabilitive (verojatnostno), which likewise displays the marked nonconfirmative complex of meanings. The village is located in the northwestern corner of Bulgaria only a few kilometers from both Romania and Serbia. The future is formed, as elsewhere in Bulgarian, with a preposed invariant marker, čă, which is the local variant of the ‘want’-based particle. The dialect also has a series of paradigms based on a Serbian-type future, i.e., conjugated enclitic ‘want’ suffixed to the infinitive stem, which now functions as a present with its own future as well as a past using the probabilitive of ‘be’ as the auxiliary.Footnote 368 The forms are illustrated in Table 6.34 (based on M. Mladenov, 1969: 109–110). Given the location of this dialect, it would appear that what happened in Novo Selo is that with the establishment of the Bulgarian-type future, the Serbian type, with which it had been in competition, was restricted to the nonconfirmative meanings found in the Romanian presumptive, which also makes use of future marking. Thus, as with the Meglenoromanian inverted perfect under Macedonian (and probably Turkish) influence and Albanian under Turkish (and in some places perhaps Macedonian) influence, a positional variant acquired evidential-like meanings.
| Indicative | Probabilitive | |||
|---|---|---|---|---|
| sg | pl | sg | pl | |
| present | ||||
| 1 | glèdăm | glèdămo | glădàčăm | glădàčămo |
| 2 | glèdăš | glèdăte | glădàčăš | glădàčătă |
| 3 | glèdă | glèdăju | glădàčă | glădàčăju |
| future | ||||
| 1 | čă glèdăm | čă glèdămo | čă glădàčăm | čă glădàčăm |
| 2 | čă glèdăš | čă glèdăte | čă glădàčăš | čă glădàčătă |
| 3 | čă glèdă | čă glèdăju | čă glădàčă | čă glădàčăju |
| past | ||||
| 1 |
|
|
|
|
| 2 | si glădàl | ste glădàli | budàčăš ~ bùdăš ~ bìčăš glădàl | budàčătă ~ bùdătă ~ bìčătă glădàli |
| 3 | e glădàl | să glădàli | budàčă ~ bùdă ~ bìčă glădàl | budàčăju ~ bùdăju ~ bìčăju glădàli |
Some examples of the use of the probabilitive are given in (6.235)–(6.244). Example (6.235) is a past reported or inferential; (6.236) is a past in which the speaker accepts that the event occurred but disavows responsibility for it; (6.237) and (6.238) are present reporteds or inferentials; (6.239) is a kind of admirative-dubitative using a past form to refer to a present state of affairs; (6.240), (6.241), and (6.242) all involve attenuated questions – future, past, and present, respectively – illustrating that the probabilitive can be used in situations when the speaker lacks information but can expect an indicative answer from the addressee; (6.243) is a present probabilitive with admirative meaning, and (6.244) is future interrogative in form, but unlike the questions in (6.240–6.242), it is basically rhetorical and expresses annoyance, hence related to dubitative meaning.
(6.235)
Budàčătă bilì i vì tàmo. be.prob.2sg be.ptcp and you there ‘You [apparently] were there, too.’
(6.236)
Kokò bìčă stănùlo nèznăm. how be.prob.3sg happen.ptcp neg.know.prs.1sg ‘I don’t know how it happened.’
(6.237)
Răbotèčă nìgdăm. work.PROB.3SG somewhere ‘He is [apparently] working somewhere.’
(6.238)
Kopàčăju săgà lòjză. dig.prob.3pl now vineyard ‘They are [apparently] hoeing the vineyard now.’
(6.239)
Kădà budačăm imàl ja tòlko părè? when be.prob.1sg have.ptcp I so.much money ‘When would I have so much money?’
(6.240)
Jutră čă kopàčăš lòjză? tomorrow fut dig.prob.2sg vineyard ‘Will you [by any chance] be hoeing the vineyard tomorrow?’
(6.241)
Dălì bičă minùlo po ovdè golămò dătè? q be.prob.3sg pass.ptcp by here big child ‘Did a big child [by any chance] pass by here?’
(6.242)
Imàčăš kurùz u koš? have.prob.2sg corn in silo ‘Do you [by any chance] have corn in [your] silo?’
(6.243)
Kokò pièčă tòlko ml’ògo how drink.prob.3sg so.much much ‘How much he drinks! / How does he drink so much?!’
6.2.5.10 Greek
Greek does not have any evidential uses of paradigmatic tense forms. It does, however, use λέει ‘one says’ as a kind of particle expressing a complex of evidential-like meanings including the expected ‘reportedly, allegedly,’ as in (6.245), but also an admirative-like emphatic, as in (6.246):Footnote 369
(6.245)
–Θα πάμε στην Πράγα; – Α! είναι, λέει, πολύ ωραία. (Grk) fut go.1pl to.the.acc.sg P. OK be.prs.3sg evd very pretty.nom.sg –Ποιός λέει; who say.prs.3sg –Όχι, είναι, είναι πολύ ωραία. no be.prs.3sg be.prs.3sg very pretty.nom.sg ‘Shall we go to Prague? –Yeah, it’s supposed to be very beautiful. –Who says? –No, it really is very beautiful.’ (César Montoliu, p.c.)
(6.246)
–Ήταν καλό το φαγητό; –Καλό, (Grk) be.impf.3sg good.nom.sg the.nom.sg meal.nom.sg good.nom.sg λέει! evd ‘–Was the meal good? –Very good!’ (Kriaras 1995)
The development of λέει into a quotative particle in Greek expressing the nonconfirmative complex of meanings illustrated in (6.245) and (6.246) is typologically interesting, since it appears to offer a start to the emergence of a grammatical marker of evidentiality, but it does not enter into the kind of paradigmatic manifestations of evidentiality seen in the other Balkan languages.
By contrast, Cypriot Greek has borrowed Turkish mIş as a dubitative particle with a native (Greek) 1st person dative/genitive marker, viz. mišimou (cf. Friedman 2018c; Kappler & Tsiplakou 2018). In the case of mišimou, only the dubitative meaning is normal, but it is still connected to markedness of first person intersecting with markedness of nonconfirmativity. Kappler & Tsiplakou 2018 give the following examples demonstrating that the particle is not limited with regard to which constituent it occurs with:Footnote 370
a.
mišimou εμίλησεν ο Γιώρκος της Σταυρούλλας dub spoke.3sg the.nom George.nom the.gen Stavroula.gen b.
εμίλησεν mišimou o Γιώρκος της Σταυρούλλας spoke.3sg dub the.nom George.nom the.gen Stavroula.gen c.
εμίλησεν o Γιώρκος mišimou της Σταυρούλλας spoke.3sg the.nom George. nom dub the.gen Stavroula.gen d.
εμίλησεν o Γιώρκος της Σταυρούλλας mišimou spoke.3sg the George.nom the.gen Stavroula.gen dub ‘George spoke to Stavroula – yeah, right!’
To these we can add an example from Marina Terkourafi (University of Leiden, p.c. 2018) with a present tense:
(6.248)
εν᾽ πλούσιος mišimou is rich dub ‘He claims to be rich (but I don’t believe it).’Footnote 371
It is the connection of nonconfirmativity to first-person marking that turns out to be relevant to the Cypriot borrowing of Turkish mIş. For the Cypriot Greek borrowing, in addition to miši mou, Kappler & Tsiplakou 2018 report the forms “miʃi, imiʃ and even miʃteti,” all of which they label as “basilectal” and the last as “on its way to full obsolescence” (see also Tsiplakou et al. in prep.).Footnote 372 We can note here that Hadziioannou 1996 also reports múši.
As Kappler & Tsiplakou 2018 note, and as other consultants confirm (Terkourafi, p.c.), the form with mou, i.e., míši mou is by far the more common. Kappler and Tsiplakou offer comparisons with Cypriot Greek περκίμου, cf. Standard Turkish belki, dialectal Turkish belkim ‘perhaps,’Footnote 373 Cypriot Greek ατšαπής μου ‘I wonder’ from Turkish acaba ‘I wonder’ and αφερίμου ‘well done!’ from Turkish aferim ‘idem,’ which is from Persian âferin (Middle Persian āfrīn ‘praise, blessing’). The examples derived from belkim and aferim have an -m in the source language, but the Cypriot copy of acaba has a completely innovative mou, and, moreover, the source form has an inherent first-person referent. To this can be added the fact that at least some Cypriot Greek speakers consider the form mîši su, with 2sg genitive pronoun su (σου), possible (Terkourafi, p.c. 2008).
A striking difference between the Cypriot Greek and the Topaanli Romani discussed in §6.2.5.3 is that while the Greek particle is strictly dubitative, in Topaanli Arli the 1st singular marker is not associated with dubitativity. This in turn points to a more general association of first person and the admirative complex in general.
6.2.5.11 Evidentiality: Summary
While the question of whether Balkan Slavic and Albanian evidentials resulted from internal development or external influence continues to be debated, the fact that Turkic shows evidential usage already in the eighth century while Albanian and Balkan Slavic do not yet have fully developed evidential systems in the sixteenth but do have them now means that even if those languages had the internal possibility of developing such systems or usages on the basis of native material prior to the Ottoman period – what Enfield 2003: 5 calls typological poise – what we know about the social position of Turkish and of multilingualism in the Balkans during Ottoman rule makes it impossible to discount Turkish influence in this respect. Only donning the blinders of nineteenth-century nationalism would allow one to ignore Turkish here. For the other languages considered in this section, the Balkan-Ottoman source is obvious in some cases, while in others, the time of development is Ottoman.
We note too that what has been described in these discussions of evidentiality and modality reveal nuances of interpretation that conversational interaction – with the full richness of context that such interaction requires – must be assumed in order to make sense of how speakers of different languages could come to converge on these details. Thus even though a matter of (morpho)syntax and pragmatics, the modality and evidentiality convergences actually fit in well with, and can be argued to follow from, the conversationally based ERIC model advanced in §4.3 for lexical convergence.
6.2.6 Voice and Valency
Verbal “voice,” also known by the label “diathesis,” is a complex notion in grammatical theory but generally refers to the grammatical encoding of the perspective on how the participants in the action expressed by a verb relate to one another. For instance, in the expression of a transitive event, in which an actor acts upon a patient, voice marking could indicate whether the action is expressed from the perspective of the actor as the prominent entity, or instead from that of the patient. Typically, the participant given prominence surfaces as the subject of the clause. Indeed, Shibatani 1988: 1 overtly builds this into his characterization of voice, seeing it “as a mechanism that selects a grammatically prominent syntactic constituent – subject – from the underlying semantic functions (case or thematic roles) of a clause.”
There are thus discourse-related issues with voice, having to do with “prominence,” but voice also pertains to argument structure and how the participants in an action are grammatically distributed in a given verbal expression. In many instances, the relationship between different voice categories centers on whether an underlyingly transitive event (X affects Y) is grammatically expressed as transitive or intransitive, and by the same token, whether an overlay of causation is added to underlyingly intransitive events giving ultimately a transitive expression. For this reason, discussions of voice necessarily touch on matters of valency, and in the typical voice opposition of active versus nonactive, the latter generally subsumes various kinds of detransitivization, while the former can have a causative sense. Thus, some of the discussion here is not on voice per se but more on matters of argument structure.
As far as the Balkans are concerned, there are parallels in form and function involving voice, but overall there is relatively little of Balkanological interest. Accordingly, two aspects of voice in the Balkans are the focus here: first, some parallelisms involving the nonactive voice, and second, voice lability, where the same forms can be both intransitive and transitive (thus causative in some instances) so that the valency of the verb is affected.
6.2.6.1 Nonactive Parallelisms
The overall structure of oppositions and functions involving nonactive forms in the Balkans is similar in all of the Indo-European branches represented, and also in Turkish. They all have a contrast between active and nonactive voice forms, where nonactive subsumes forms – and functions – that are variously referred to as “middle,” “mediopassive,” or simply “passive,” but which, for the Indo-European languages at least, actually take in a wide range of functions, including some intransitives with no active counterpart (so-called “deponent” or “medium tantum” verbs), reflexives, and even reciprocals. Moreover, nonactive voice forms are generally intransitive (whereas active voice forms are either intransitive or transitive).Footnote 374 Examples of these different types include the synthetic forms in (6.249):
a.
Middle: Alb merzitem ‘I get bored,’ dëshpërohem ‘I become disappointed’ b.
Passive: Alb forcoheni (prej profesorit) ‘you are forced (by the professor),’ Grk ακούστηκαν (απ᾽ όλο τον κόσμο) ‘they were heard (by everyone)’ c.
Medium tantum: Alb kollem ‘I cough,’ Grk κοιμούμαι ‘I sleep’ d.
Reflexive: Alb krihem ‘I comb myself,’ lahem ‘I wash myself,’ Grk χτυπιέμαι ‘I hit myself’ e.
Reciprocal: Alb takohemi ‘we meet one another,’ Grk συναντιόμαστε ‘we meet one another’
This superficial parallelism is likely in this case to be the result of inheritance, or possibly independent development, because all of these are features of so-called middle voice inflection, reconstructible for Proto-Indo-European and found across the Indo-European family, in some instances with innovative material, as with Romance and Slavic verbs using se (or the historical equivalent thereof). There is thus an Indo-European element to the commonality here, rather than a specifically Balkan one. As with other such parallels, this coinciding of form and function can be seen as contributing to a sense of structural commonality across the languages, but without a contact-related basis for it.
However, there is a parallel involving nonactive voice in Romani and Macedonian that is likely to be the result of contact, and that thus has Balkanological relevance. Most of Romani has a Romance-type reflexive conjugation, such that first and second persons use a personal pronoun (thus, “agreeing” with the subject) rather than the reflexive pronoun. Thus, for example, ‘urinate’ – Romani mutrav, Macedonian moča – is a transitive verb. The usual expression of the bodily function is an intransitive nonactive form; in Romani, e.g., in the 1sg form, this is mutrav man with the agreeing pronoun man ‘me,’ whereas in Macedonian the form is se močam, with invariant (nonagreeing) se ‘self.’ However, some Romani speakers in North Macedonia have calqued the Macedonian usage and use the invariant reflexive pronoun pe(s) in this nonactive formation. As a result, some speakers say džav(a) te mutrav pe, with mutrav pe like Macedonian se močam, rather than džav(a) te mutrav man for ‘I’m going to take a pee.’ To speakers of more conservative Romani dialects, this former expression means ‘I’m going to go pee on myself.’
Finally, with regard to the form of nonactive voice expression, Greek and Albanian coincide in two ways that are intriguing but in the end perhaps Balkanologically inconsequential.Footnote 375 In particular, in both of these languages, there are special endings in the present systemFootnote 376 for signaling nonactive voice, endings which are distinct from the set of active endings; this situation is illustrated by these present tense paradigms for ‘wash’ (i.e., nonactive: ‘I am washed / I wash myself,’ etc.; active: ‘I wash,’ etc.); see Table 6.35.
| Albanian Nonactive | Greek Nonactive | |||||||
|---|---|---|---|---|---|---|---|---|
| 1sg | lahem | 1pl | lahemi | 1sg | πλένομαι | 1pl | πλενόμαστε | |
| 2 | lahesh | 2 | laheni | 2 | πλένεσαι | 2 | πλένεστε | |
| 3 | lahet | 3 | lahen | 3 | πλένεται | 3 | πλένονται | |
| Albanian Active | Greek Active | |||||||
|---|---|---|---|---|---|---|---|---|
| 1sg | laj | 1pl | lajmë | 1sg | πλένω | 1pl | πλένουμε | |
| 2 | lan | 2 | lani | 2 | πλένεις | 2 | πλένετε | |
| 3 | lan | ~ | 3 | lajnë | 3 | πλένει | 3 | πλένουν |
Moreover, in the simple past (aorist) tense, a different, but still parallel, strategy is to be found for marking nonactive voice forms: active endings are added to a stem that is characterized in a special way for being nonactive: in Greek the formative -(θ)ηκ- is suffixed to the past perfective stem and in Albanian the formative u is prefixed to the past active stem (aorist, optative, admirative, participle), as in the forms for ‘I was washed, I washed myself,’ etc. (nonactive), ‘I washed’ (active) in Table 6.36, with the similar endings highlighted (italics in Albanian, underlining in Greek).Footnote 377
| Albanian | Nonactive | Active | Greek Nonactive | Active |
|---|---|---|---|---|
| 1sg | u lav-a | lav-a | πλύθηκ-α | έπλυν-α |
| 2 | u lav-e | lav-e | πλύθηκ-ες | έπλυν-ες |
| 3 | u la-Ø | lau-Ø | πλύθηκ-ε | έπλυν-ε |
| 1pl | u la-më | la-më | πλυθήκ-αμε | πλύν-αμε |
| 2 | u la-të | la-të | πλυθήκ-ατε | πλύν-ατε |
| 3 | u la-në | la-në | πλύθηκ-αν | έπλυν-αν |
What gets in the way of this striking parallelism being of significance to Balkan linguistics is the fact that Ancient Greek showed a similar set of facts, with special endings in the present and active endings with a special formative in the past, illustrated here just with singular forms (Table 6.37).
| Present | Past | |||
|---|---|---|---|---|
| Nonactive | Active | Nonactive | Active | |
| 1sg | πλύνομαι | πλύνω | ἐπλύθη-ν | ἔπλυνο-ν |
| 2 | πλύνῃ | πλύνεις | ἐπλύθη-ς | ἔπλυνε-ς |
| 3 | πλύνεται | πλύνει | ἐπλύθη-Ø | ἔπλυνε-Ø |
Thus either there was contact leading to the convergence in ancient times, or else this could possibly be a basis for positing a Greek-Albanian dialect grouping within Indo-European.Footnote 378 Either possibility may be of interest in itself but not to the later Balkan sprachbund.
However, it is conceivable that even if inherited into both Albanian and Greek, these patterns were reinforced in each language through contact, as it is a striking fact that among all modern Indo-European languages, only these two in the Balkans preserve the inherited feature of special nonactive presential endings. Moreover, the Albanian use of an invariant u – invariant in that it does not change with the person of the subject – is quite like the Slavic use of invariant se (from earlier sę), based ultimately on the Proto-Indo-European reflexive stem *swe-,Footnote 379 which represents the Slavic innovative strategy for encoding nonactive forms distinct from active. Admittedly, the Slavic strategy differs from the Albanian in that se is not restricted to the nonpresent tense categories the way that u is in Albanian, but focusing just on the invariance offers a possible point of contact influence between Slavic and Albanian, and given that the Slavic phenomenon is in fact Balto-Slavic, this may be another feature relevant to the claim of a Northwest Indo-European contact period for Albanian. Thus in the course of later interactions and contact, Greek and Albanian could have mutually reinforced one another regarding the choice of nonactive marking strategy and Slavic may have contributed to the Albanian invariance of u.Footnote 380 Still, there are enough uncertainties in this account that its relevance for the Balkan sprachbund must remain speculative.
6.2.6.2 LabilityFootnote 381
As suggested above in §6.2.6, lability refers to the use of the same voice-indifferent form for both intransitive and transitive – typically causative – events, in a system where nonactive voice might be expected for the intransitive. Within the Balkans, lability is a feature found in Balkan Romance, Albanian, Balkan Slavic, and Greek, these last three being languages with other fairly well developed voice-marking (detransitivizing) strategies. In Balkan Slavic, especially Macedonian, lability shows the effects of contact, and sets it apart historically from the rest of Slavic. At the same time, Romani, a language with a well-developed causative strategy, seems to avoid lability, so that lability seems to be both contact-induced and contact-resistant in the Balkans. One important consequence of the examination of Balkan lability is that Macedonian and Aromanian turn out to be more similar to one another than to their Balkan genealogical relatives Bulgarian and Romanian, respectively (Gołąb 1970).
The relevant kind of lability for the Balkans is what Kulikov & Lavidas 2014: 872 call Patient-preserving lability (P-lability), where the transitive expression has a causative sense, as in (6.250), where (a) means essentially ‘John caused (6.250b),’ i.e., ‘John caused the vase to break’:
a.
Ο Γιάννης έσπασε το βάζο. the.nom.sg John.nom.sg break.pst.3sg the.acc.sg vase.acc.sg ‘John broke the vase.’ b.
Το βάζο έσπασε. the.nom.sg vase.nom.sg break.pst.3sg ‘The vase broke.’
The particular P-lability of (6.250) does not occur in Balkan Slavic, Balkan Romance, or Albanian with ‘break,’ which would require a nonactive form for (6.250b); such a causative sense does occur, however, in Balkan Slavic and Balkan Romance with ‘burn,’ as in (6.251) from Bulgarian and (6.252) from Aromanian, where the sense is ‘X causes the trees/village to burn’:
a.
Dărvata gorjat (Asenova 2002: 109) trees.the burn.3pl ‘The trees burn’ b.
Dărvata gi gorjat. (Asenova 2002: 109) trees.the them.acc burn.3pl ‘They burn the trees.’
a.
Hoara ardi baba s-cheaptinã. (Cuvata 2009: 65) village.the burns.3sg grandma refl-combs.3sg ‘The village burns, grandma combs her hair.’ [i.e., Nero fiddles while Rome burns.] b.
Andartsilji arsirã Avdhela tu 1905. (Papahagi 1974: 194) rebels burned.3pl Avdhela in.the ‘The Greek bandits burned Avdhela in 1905.’
For these last examples, in Macedonian and Romanian the same verb would also be used, e.g., Macedonian gorat in (6.251), Romanian arde and arsirã in (6.252), respectively (cf. also Koneski 1981: 181); in Albanian, however, the intransitive in each case would require a nonactive form, e.g., drutë digjen ‘The-trees burn,’ versus transitive drutë djegin ‘They-burn the-trees’ (= ‘They-cause-to-burn the-trees’). In the case of Romani and Turkish, etymologically related but synchronically distinct verbs in each language are needed for intransitive and transitive ‘burn’: intransitive phabol and yan-, transitive phabarel and yak-, respectively.
Other Slavic languages would not allow such P-lability with their equivalent (i.e., underived imperfective) forms of ‘burn,’ thus localizing the lability in the Balkans as far as Slavic is concerned. Still, there are differences in P-lability between Bulgarian and Macedonian, in that there are labile verbs in Macedonian whose Bulgarian counterparts are not labile:
a. Umram.
‘I die.’
a’.
Go umram čovekot. him.acc die.1sg man.the ‘I kill the person.’ b. Zaspivam.
‘I fall asleep.’
b’.
Go zaspivam deteto. him.acc sleep.1sg child.the ‘I put the child to sleep.’ c. Šetam.
‘I walk.’
c’.
Go šetam kučeto. him.acc walk.1sg dog.the ‘I walk the dog.’
For (6.253a’), Bulgarian would have to use ubie ‘kill,’ which is possible also in Macedonian, but only Macedonian can use ‘die’ transitively to mean ‘cause to die.’ For (6.253b’), the meaning of zaspiva is ‘fall asleep’ in both languages, but only in Macedonian can the verb also mean ‘lull to sleep’ (i.e., cause to sleep); for this last meaning, the overtly causative prispiva would have to be used in Bulgarian. Finally, as for (6.253c) and (6.253c’), there are both lexical and valency differences between Macedonian and Bulgarian. In Bulgarian, šeta means ‘do housework’ or ‘roam,’ and the Bulgarian equivalents of Macedonian šeta are razxožda (transitive) and razxožda se (intransitive). The Macedonian equivalent for these last Bulgarian verbs, se razoduva, means ‘begin to walk’ and, like the Bulgarian, is not labile; only Macedonian has the option of using šeta as a labile verb.
As for Balkan Romance, as noted above, Aromanian patterns with Macedonian, while Romanian patterns with Bulgarian. Thus Aromanian mor(i), do(a)rme, and (pri)imnã have the same lability as umre, zaspiva, and šeta, respectively in Macedonian, whereas Romanian muri, dormi, and plimba do not (the first two are intransitive, the last is transitive and must be marked with se in order to be intransitive, as in Bulgarian).
In Greek, as shown by (6.250) above, lability occurs too, and other examples with motion verbs can be adduced:
a.
Πήγε στο σχολείο (Asenova 2002: 109) go.pst.3sg to.the.acc.sg school.acc.sg ‘He went to school.’ a’.
Τον πήγε στο σχολείο (Asenova 2002: 109) him.acc go.pst.3sg to.the.acc.sg school.acc.sg ‘He took him to school.’ b.
Θα περπατήσω στον κόσμο της νύχτας fut walk.1sg into.the.acc.sg world.acc.sg the.gen.sg night.gen.sg ‘I will walk into the world of the night’ b’.
Θα έρθεις να σε περπατήσω στον κόσμο fut come.2sg dms you.acc walk.1sg into.the.acc.sg world.acc.sg της νύχτας;Footnote 382 the.gen.sg night.gen.sg ‘Will you come so that I might walk you into the world of the night?’ c.
Ταξιδεύω παντού travel.1sg everywhere ‘I travel everywhere’ c’.
Γιατί ξέρουμε να σας ταξιδεύουμε παντού!Footnote 383 because know.1pl dms you.acc travel.1pl everywhere ‘Because we know how to “travel you” everywhere!’
Even with such examples, there are lexically specific constraints on, as well as variation concerning, lability in Greek. For instance, intransitive ‘sleep’ is κοιμούμαι while transitive ‘put to sleep’ is the (historically related but nonetheless distinct) verb κοιμίζω; intransitive ‘die’ is πεθαίνω while transitive ‘kill, cause to die’ is σκοτώνω, though interestingly πεθαίνω can be used transitively in a metaphorical sense, e.g., Με πεθαίνει η δουλειά ‘Work is killing me!’ (literally, ‘me.acc kills the-work.nom) but not in a literal sense, e.g., *Τον πέθανε η Μαφία ‘The Mafia killed him’ (literally, ‘him.acc killed the Mafia.nom).Footnote 384
Turkish has a well-established causative system, with several suffixes, most productively -DIr-, that derive causatives from basic roots. It is thus not surprising that no lability occurs in Turkish, not even in the West Rumelian dialects. A similar observation is to be made for Judezmo, which entered the Balkans with a well-developed Romance-style diathesis system, with detransitivized formations utilizing reflexive pronouns , as in asentar ‘seat (someone)’ ~ asentarse ‘be seated; seat oneself,’ and shows no lability.
Romani, however, offers a different situation. Historically, in Romani there has been a robust system of causative formation, albeit to various degrees in various dialects, and this system has been transformed.Footnote 385 An inherited causative formant is -av-, as in (6.243b):
a. Me phirava.
‘I walk.’
b.
Me phirava-va e džukele. I walk-caus the dog ‘I walk the dog.’
Nonetheless, some recent loanwords into Skopje Romani show the lability of the Macedonian source verb. One such loan is plivinel ‘to swim,’ from Macedonian pliva:
a.
Plivam (Macedonian) = Plivinava (Romani) ‘I swim.’ b.
Go plivam kučeto. (Macedonian) = Plivinava e džukele (Romani) it.acc swim.1sg dog.the swim.1sg the dog ‘I give the dog a (chance to) swim.’
The picture that emerges from a consideration of these facts about lability in the Balkans is that, as with other pan-Balkan features, e.g., the infinitive (Joseph 1983a, and §7.7.2.1.1) or the Dental Modal Subordinator (“dms,” cf. Friedman 1985a, and see §4.3.3.1.2, Footnote footnote 145, and §7.7.2.1.3.1), lability is a scalar Balkanism. At the far end of nonlability there is Turkish, with Albanian and Romani close to that. Still, they are less labile than Greek, Balkan Slavic, or Balkan Romance, though in different ways. Albanian has a strong nonactive voice-marking system, so that a detransitivized pattern, as with ‘burn’ in the equivalent of (6.252), calls on nonactive morphology, whereas Romani maintains causativity and even created new causative marking (see Footnote footnote 385). On the other hand, contact with Macedonian has introduced some lability into Romani, as in (6.256ab). Within Balkan Slavic and Balkan Romance, Macedonian and Aromanian pattern together as the most innovative, whereas Bulgarian and Romanian pattern together as more conservative, thus in each case repeating clustering seen with other Balkan features. Makartsev & Wahlström 2020 provide interesting data from which they conclude that lability (“ambitransitivity” in their terms) becomes stronger as one moves from north to south, especially for Macedonian and Albanian. Their conclusions argue that Bulgarian is an outlier but Aromanian is closer to Romanian. Their study of Balkan lability is still a work in progress, but contact certainly seems to be involved in at least some of the developments.
Open access