Introduction1
The oldest attested Kurdish texts date back to the 15th century (Omarkhali, 2013). Ground breaking changes, however, occurred in Iranian languages during the transition of Old to Middle and of Middle to New Iranian. Historical landmarks are the fall of the Achaemenid Empire for the first (4th c. BC), and the fall of the Sasanian Empire for the second transition period (7th c. AD). How shall we account for the historical evolution of a language which is not attested in these relevant periods?
This paper attempts to set up a chronology of grammatical changes for several languages which – some more and some less commonly – are considered Kurdish. This will be done in comparison with attested Middle Iranian languages (see the last section). The chronology will allow us to determine when the varieties under investigation must have become distinguishable from one another with respect to a particular grammatical change. Whether this distinction separates dialects or languages is a question that will not be addressed here (see the introduction to this volume). Needless to say, language classification is much more complex than the cautious approach of this study, and involves taking all linguistic parameters into account such as the phonological system, the lexicon, etc.2
Kurdish is used in this study as a more geographically or culturally defined broad term. This does not imply that all languages discussed in this article belong to one definable linguistic unit. They happen to be spoken in, or close to, an area which is called Kurdistan. Regardless of the speakers’ identity, a linguistic definition of “Kurdish” would follow specific linguistic parameters. Dialects may be seen as closely related varieties of a language, while a sprachbund (area of linguistic convergence) is a bundle of areally contiguous languages which may or may not be genetically related, but share specific features. However, whether we should speak of Kurdish dialects or of a Kurdish sprachbund, or whether we should abandon Kurdish as a linguistically defined term altogether, is not relevant for the aim of this study. In the same way, terms such as “Kurmanji” refer to the corpora I took the data from, while ignoring dialectal variation within “Kurmanji” itself (see Haig and Öpengin in this volume). In order to avoid confusion, I will refer to Kurmanji as “Botan-Kurmanji” and to Sorani as “Sılēmani-Sorani” whenever necessary. Information on the respective varieties were mainly gathered from the following sources: Gorani (Mahmoudveysi et al., 2012; Mahmoudveysi and Bailey, 2013), Hawrami (MacKenzie, 1966), Kurmanji (Bedir Khan and Lescot, 1970), Sorani (MacKenzie, 1961a), Southern Kurdish Dialects (Fattah, 2000), and Zazaki (Paul, 1998a and 2009).
On Classifying Iranian Languages
The traditional classification of Iranian languages, which is still widely used, follows the tree model, i.e., Proto Iranian is divided into two groups (Proto West and Proto East Iranian), and then further into Proto Northwest and Southwest, and Proto Northeast and Southeast Iranian. This classification is mostly based on observations of Middle Iranian languages, e.g., accent shift and sonorisation of voiced plosives. The accent shift in Middle West Iranian languages resulted in the loss of endings, while the Middle East Iranian languages usually retained them.3 Voiced plosives become fricatives or approximants in all positions except for word initial position in West Iranian.4 However, neither feature yields a clear-cut division of West and East Iranian. The “heavy stems” of the East Iranian language Sogdian also show a loss of endings and only two differing case forms for nouns: direct and oblique. This subsystem corresponds to the earlier stages of Middle Persian and Parthian (see the section on the nominal system). However, several case systems of New West Iranian languages can be related to Old Iranian case forms (e.g. Cabolov, 1997: 54). Hence, the loss of final syllables in West Iranian did not affect all endings in every West Iranian language. Likewise, the shift of voiced plosives to fricatives or approximants does not occur in a uniform manner in the attested languages, cf. the development of /g/ in the following table.
Development of voiced plosives in Middle Iranian
In the initial position West and East Iranian are clearly separated. In the intervocalic position, it is only Middle Persian that differs. We may describe the sonorisation process as a scale with its strongest effect in the Southwest (Persian g > *
The tree model, which assumes discrete and absolute separation between its branches, is obviously not well-suited as a representation of the kind of overlapping features previously discussed. Serious problems also occur if the classical tree model is applied to a division of Old Iranian languages. Since the dichotomy of West and East Iranian is mainly defined by developments that took place during the transition of Old to Middle Iranian, its criteria are not valid for Old Iranian languages.6 There has been considerable debate regarding the position of Avestan vis à vis the East/West distinction, and with the discovery of Bactrian documents in the 1990s, it became clear that Bactrian occupies an intermediate position.7 Such observations, among others, led Sims-Williams (1996) to propose a language continuum and for Old Iranian a division into a centre and a southern and northern periphery. Figure 1 gives a schematic representation of this view (adapted from Jügel, 2013: 301).
Combining the tree model with areal phenomena
The Old Iranian trichotomy can be represented by the tree model. The inter-relationship of Old Iranian varieties should be understood as one of dialects in a continuum with the peripheries being the continuum’s furthest anchors. From the Old Iranian period onwards, migration and areal phenomena superimpose on the older division in varying ways.
When modern varieties are linked to the Old Iranian stage, factors like migration and areal phenomena need to be considered in order for the representation to reflect the linguistic complexity of each variety’s history (and not merely a division according to a few selected isoglosses). At first, it is best to take each variety individually and to trace its development in a dynamic scheme. In a second step, one can identify those varieties which share so many features that they can be considered the continuation of a particular variety through time, hence linked by a single line, cf. Fig. 2.
The Old Iranian trichotomy is represented by SP (Southern Periphery), Centre, and NP (Northern Periphery). The shaded bars represent contact areas or the transition zones in the sense of language continua. In reality, these zones will have moved, expanded, or shrunk over time. Each box stands for one variety. The circles group varieties together. If the varieties can be identified as dialects, one can interpret such groups as languages. If the varieties represent languages, these groups may be called sprachbunds. In the figure above, only two varieties remain in one group, while other varieties come and go. Consequently, these two varieties might be understood as dialects of one language, which was part of differing sprachbunds in the course of its development (so the circles represent sprachbunds here). The chart can be expanded by the representation of sub-, ad-, or superstratum, e.g., the influence of Persian. Differences in the intensity of influence can be represented by shorter or longer arrows and paler or darker colouring, as in Fig. 3.
Representation of sub, ad-, or superstratum
The Kurdish development could be represented as in Fig. 4. An assumed Proto Kurdish split into several varieties which came under areal influence with others (represented by the two circles). The variety in the intermediate position is historically the intermediate variety of the Kurdish dialect continuum and nowadays a contact area of the two sprachbunds. Instead of one Proto Kurdish language (Fig. 4a), we could also assume an original Kurdish sprachbund (Fig. 4b). In addition, one of the Kurdish varieties may in fact be an older lingua franca, which became the native language of a specific region.
MacKenzie (1961b) pictured the development of Kurdish quite differently. According to him, Kurdish moved northwest and came into contact with Zazaki. From the northern position it spread south and “overran” Gorani speaking territory. The absorption of Gorani led to the deviation of Central and Southern Kurdish, while Northern Kurdish, i.e., Kurmanji, preserved to a much greater extent its “purity” (p. 86). This scenario seems to be primarily based on two assumptions, viz. Kurmanji is prototypical Kurdish, and the Gorani “speech islands in a sea of Kurdish” (p. 73) are the remnants of an earlier contiguous Gorani speaking area. However, language communities may split and migrate, and whether all Kurdish varieties are related to one language requires further investigation. If Kurmanji spread into Gorani speaking territory and differences among Sorani and Kurmanji are due to the Gorani substratum, it is hard to explain why today’s Sorani does not have morphologically marked case, because today’s Kurmanji and Gorani still preserve it.
Finally, one last issue will be addressed here. In a continuum of varieties, each variety differs from its immediate neighbours through small differences. But these small differences accrue across a chain of varieties, so that the peripheries of the continuum differ from each other quite significantly. Over time, these peripheral varieties may shift their locations through migration, and then come into contact with each other. In such a scenario, the two peripheral varieties may then influence each other, and we will find the kind of language phenomena that are typical of any contact situation involving two distinct languages. Thus we find a combination of a dialect continuum, overlaid with a secondary region of language contact, which can be further complicated by contact from neighbouring languages. If we hypothetically transfer this model onto Kurmanji and Sorani in order to make it more concrete, some areas in between Kurmanji and Sorani might be contact areas, while in other areas we could detect a continuum of gradual dialectal changes, assuming these two varieties have a common ancestor at all. In the next sections, I will identify and evaluate selected grammatical features against the background of the model sketched above.
Grammatical Features
I will mainly focus on the following grammatical features: grammatical gender, case, and article systems in the nominal system; verbal agreement, verbal stems, and encoding patterns of clausal arguments (e.g., object marking) in the verbal system.
Nominal System
Proto Old Iranian had seven cases (eight with the vocative), and three grammatical genders (masculine, feminine, and neuter). In Middle Iranian we usually see the result of a reduction of the system. For West Iranian languages we can reconstruct a two-case system: an unmarked direct case, and a marked oblique case. Together with number (singular and plural) we find the following systems among others (cf. Stilo, 2009 for New Iranian).
This schema, taken from Jügel (2012/I: 169), displays the most common systems that can be found in the Middle Iranian languages such as Bactrian, Parthian, and Middle Persian. While types 1–3 allow for a distinction of case and number, type 4 displays a pure number system.
As a rule, the oblique case sg. shows a palatal vowel.8 It is most likely derivable from the genitive singular of the a-stems, i.e., *-ahya.9 The plural (sometimes only used as an oblique plural as in types 2 and 3) goes back to the genitive plural of the a-stems, i.e., *-ānām.10 In Middle Persian and Parthian the oblique singular was lost11 and the oblique plural was generalised as a plural marker, regardless of case. The only remnants of oblique case were the enclitic pronouns and a few forms of free pronouns. In Middle Persian, these are an for the direct case, and man for the oblique case of the first person pronoun. The direct case is only attested in the inscriptions, early Manichaean texts, and in a few quotations of older text in Zoroastrian Middle Persian (Jügel, 2012/I: 220). In Parthian, the case distinction of the 1st singular is maintained throughout the corpus: az for direct case, man for oblique case.12 With respect to gender, we can only state that the distinction was lost long before our first attested Middle Persian and Parthian texts were composed. This is foreshadowed by the sporadic coalescence of declension classes in late Old Persian and Young Avestan. Pronominal case is retained in Parthian longer than in Middle Persian. However, both preserve the enclitic pronouns as oblique forms.
In New Western Iranian, morphological case systems can be found, roughly speaking, more in the north, while the Southern varieties tend to abandon case with the exception of some Gorani varieties. Case and gender distinction can be observed in Botan-Kurmanji and Zazaki. Pronouns have two forms. The 1st and 2nd persons show an inherited suppletive paradigm (e.g., 1sg. ez vs. min13), while the 3rd persons follow the nominal inflexion with suffixed case endings. In this respect, Kurmanji, Hawrami, and Zazaki are more archaic than Middle Persian and Parthian. Southern dialects of Zazaki developed a special oblique ending with a “kinship-r” (Paul, 2009: 548), which goes back to the oblique forms of Middle Iranian kinship terms, e.g., Middle Persian direct singular brād vs. oblique singular brādar “brother”,14 Zazaki direct singular bırā vs. oblique singular bırār “brother”.15 In contrast to the attested Middle West Iranian languages, Kurmanji and Zazaki have lost enclitic pronouns.
There are Iranian varieties which exhibit pronominal and nominal case distinction as well as enclitic pronouns (e.g., Taleshi, Tati, Vafsi), which is also true for Mukri and Hawrami with the exception of the 1st and 2nd persons of the free pronouns, which are invariant (MacKenzie, 1966: 24; Öpengin, 2013: Sections 2.3.2.5.3 and 2.3.5.1.1).16 Central and Southern Sorani, the Gorani language of Gawraǰū,17 and the Southern Kurdish Dialects, on the other hand, have lost case marking completely and conform to New Persian in this respect.
The marking of grammatical gender yields a similar picture. Kurmanji retains gender as well as case, though gender is only visible within ezafe (ez āfe) constructions and in the two distinct forms of the singular oblique case, which are still preserved in most dialects of Kurmanji (Bedir Khan and Lescot, 1970: 104). According to MacKenzie (1954: 537), the gender distinction in Kurmanji is inherited. The same would also hold true for Hawrami, Zazaki, and others.18 Again, languages in the Southeast (except for Hawrami) do not show grammatically defined gender. There are differences in the vocative in Sorani (MacKenzie, 1961a: 57), but they are defined by sexus.19
Languages which lie more to the South on our scale exhibit another distinctive feature: an article system, which is a rather exceptional development for an Iranian language.20 Sorani has a fully marked system: zero marking, definite sg. -aka and pl. -akān, indefinite sg. -yak and pl. -ān (with phonological variation depending on the final segment of the stem). Also Hawrami, Gorani, and the Southern Kurdish dialects make use of definite and indefinite articles. Considering their geographical position, it is likely that Semitic languages like Arabic and/or Aramaic had their share in this development. Since we do not have historical data for the development of an article system in Western Iranian languages, it is impossible to locate this development in time.
Verbal System
Kurmanji and Sorani differ in three major fields of the verbal system: in the expression of future events, of passiveness, and in the alignment in the past domain. For the passive voice, Kurmanji makes use of the auxiliary hatin “to come” and connects it with the infinitive, e.g., ez hatim girtin “I was taken” (lit. “I came to take” or “I came to be taken”21). In this respect Kurmanji resembles New Persian, where the auxiliary šodan “to become (originally: to go)” is used together with the past participle, e.g., man gerefte šodam “I was taken”. On the other hand, Zazaki, Gorani, Hawrami, and Sorani display a synthetic way of passivisation or intransitivisation by means of stem-forming suffixes: Zazaki -i(y)-/-ey-, Hawrami -i-, Gorani (Zarda) -y- and (Gawraǰū, past) -īs-,22 Sorani -r-, Southern Kurdish Dialects -y- or -r-.23
Future sense can generally be conveyed by the present indicative. In the north, we also find specific particles with verbs in present subjunctive. Kurmanji uses the enclitic particle -ē, besides dē,24 e.g., ez-ê bikevim “I will fall” (see Unger, this volume, on the future markers of Kurmanji). This is similar to Zazaki, where the particle do is used, e.g., tı do šırē “you will go”. Sorani, Gorani, Hawrami, and the Southern Kurdish Dialects do not have a distinct future expression.25
A more complex feature is the alignment of core arguments in past tense constructions of transitive verbs. This affects the marking of direct objects and of verbal agreement. Kurmanji and Zazaki usually display ergative constructions, i.e., the logical subject appears in oblique case, the logical object in direct case, and the verb agrees with the logical object.26 However, several dialects show mixed patterns like double oblique constructions. Either subject and object show both oblique case marking, or the logical object receives an adpositional marking, comparable to the New Persian marker of the definite direct object -rā. Furthermore, verbal agreement can become opaque. These changes point to a shift from ergative constructions towards accusative constructions with marked logical objects. Such a development can be tracked in younger texts of Middle Persian (cf. Jügel, 2012/I: 452ff.). Sorani, Gorani, and Hawrami show a totally different development. The use of enclitic pronouns became obligatory for logical subjects in the past of transitive verbs. They could occur together with a coreferential noun, forming what one may call “topic agreement”. Such topic constructions are attested in Middle Iranian as well (Jügel, 2012/I: 261ff.). Hawrami displays agreement of logical object and verb, and of logical subject and enclitic pronouns, cf. ex. 1. The logical subject and object can both be expressed additionally, cf. ex. 2.
Whether the relation of the object with personal endings still is truly verbal agreement and the relation of the subject with enclitic pronouns truly topic agreement, remains to be investigated. The function of the various indexing patterns might have changed over time. In Sılēmani-Sorani and Gorani, only the former topic agreement of logical subject and enclitic pronouns remains. Logical objects (as well as any other oblique constituent, at least in Sorani) can only be indexed by personal endings if they are not expressed nominally (Jügel, 2009: 148),27 i.e., either tō-m nārd or nārd-ım-īt (dialectal nārd-īt-ım) “I sent you”. The obligatory appearance of the (morphologically speaking) enclitic pronouns is identical with the obligatory appearance of agreement markers in non-past constructions (and in past constructions of intransitive verbs). In both cases, the subject can be encoded by noun phrases, free pronouns, or it can be dropped. Confer the following examples taken from Haig (2008):
In ex. 3, the personal ending -im agrees with min, in ex. 4, it is the enclitic pronoun =m. Hence, subjects of intransitive as well as transitive verbs agree with person markers in past tense.28 The difference lies in the form of the agreement markers and in their different position in the clause.29 Since subjects of intransitive as well as transitive verbs pattern alike, while objects differ, one can conclude that transitive verbs pattern accusatively. Southern Kurdish dialects seem to make no difference between present and past tense constructions and always pattern accusatively (Fattah, 2000: 284f.).
The emergence of a specific future expression is typical for the Northern area, as is the preservation of ergative constructions. The Southern area is distinguished by the change of agreement patterns. The passive formation does not provide an obvious basis for classification: Zazaki, Gorani, and Hawrami seem to preserve an old formation, which does not, however, permit a closer grouping of these languages.
Distinctive Features of Zazaki, Gorani, and Hawrami
The flaws of our narrow perspective become immediately apparent when one takes a closer look at Zazaki, Gorani, and Hawrami. If we check the features discussed so far, we notice that Zazaki agrees in many respects with Kurmanji. Both of them retained case and gender distinctions, and both of them lost enclitic pronouns. The ergative construction is generally preserved in Zazaki, as it is in Kurmanji. On the other hand, Sorani, Gorani, and Hawrami share the article system and the peculiar agreement patterns in the past of transitive verbs. Thus, we may erroneously conclude that Zazaki and Kurmanji on the one hand, and Sorani, Gorani, and Hawrami on the other, are more closely related. In fact, this conclusion is incorrect as it ignores other features which clearly separate Zazaki, Gorani, and Hawrami from Kurmanji and Sorani.
The stems of Zazaki are tense stems just like in most of the Iranian languages, but the present stem is derived from a participle, i.e., from a diachronic perspective, it is a nominal form (Gippert, 2009: 92). In fact, we find gender distinction in the 3sg. and in some dialects also in the 2sg. of the verbal endings even in present tense, e.g., kenē “you (m.) do” vs. kenā “you (f.) do”, and keno “he does” vs. kenā “she does” (Paul, 1998a: 84). This characteristic of Zazaki, which reminds us of Semitic languages, is highly uncommon for an Iranian language.30 The demonstrative pronouns of the far deixis are identical in form with the personal endings of the 3sg.: o “that one” (m.), ā “that one” (f.). It is possible that the gender distinction in the 3sg. results from the development of demonstrative pronouns as agreement markers. This could also be true for the 2sg. if we can postulate a demonstrative with deixis to the place of the addressee (the so-called “you-deixis”) as its source. Alternatively, we may assume that pronouns became the stem of the copula.31 The difference in gender would be explained by the pronominal part of the ending, the difference in person by the copula. The reanalysis of pronouns as agreement markers is attested for a number of Iranian languages.32 This feature clearly separates Zazaki from the other varieties discussed here. However, the grammaticalisation of participles as verbal stems could have happened at any given time. Gippert (2009: 96) considers the assumption of a younger areal phenomenon possible.33 There are similar formations in several non-Iranian varieties of the region as well, in New East Armenian, and Syriac Aramaic (l.c.). Neither Hawrami and Gorani, nor Kurmanji and Sorani would have belonged to that sprachbund.
Hawrami is very archaic with respect to the verbal stems. While most of the Iranian languages reinterpreted the Old Iranian aspect stems as tense stems,34 Hawrami still uses the inherited imperfective stem to form the present and imperfect tense, e.g., 1sg. present indicative kar-u (MacKenzie, 1966: 37), 1sg. imperfect indicative kar-ene (p. 38). The other past tenses have the historical verbal adjective as a base like other Iranian languages, e.g., 1sg. past indicative kardā(ne) (p. 38), 1sg. perfect indicative kardanā (p. 39). When we look into our historical corpus of West Iranian languages, we still find reflexes of the Old Iranian imperfect in the Middle Persian inscriptions (Skjærvø, 1997; Henning, 1958: 100ff.). In later texts, no morphologically distinct imperfect forms can be found, but it seems possible that the present tense was also used as an imperfect (Jügel, 2012/I: 84ff.). So, the verbal form “present tense” still bore aspectual functions and could be used as an imperfective which was indifferent to tense.
Hawrami as well as Gorani show a specific imperfective prefix: m(ı)- in Hawrami, ma- in Gorani.35 In Hawrami its use is limited to some verbs in present tense (MacKenzie, 1966: 37), while in Gorani present as well as imperfect tense are formed with this prefix. This is in contrast to Kurmanji, and Sorani, where we find prefixes with a dental: Kurmanji di-, Sorani de- (besides e-). Most of the Southern Kurdish dialects do not use a prefix at all (as with Zazaki), though some show prefixes in the present tense similar to Kurmanji, Sorani, and Gorani: a-, (-a) ma-, dı- (Fattah, 2000: 371ff.).
Summary and Evaluation
Table 3 combines most of the discussed features of Gorani, Kurmanji, Sorani, Southern Kurdish, and Zazaki. What catches the eye is that Kurmanji and Zazaki on the one hand, and Sorani, Southern Kurdish, and Gorani on the other, cluster together in many respects.
We have seen in the previous section that Zazaki and Gorani show specific features which clearly separate them from the other varieties under investigation. Thus the grouping of Kurmanji with Zazaki, and of Sorani and Southern Kurdish with Gorani, can be taken to be an areal phenomenon. If we consider varieties between Kurmanji and Sorani, we form the impression that they share features of both. For instance, Mukri shows gender and case distinction as does Kurmanji, and enclitic pronouns and an article system as does Sorani (MacKenzie, 1961a: 50ff., 57, 76).
Do such varieties emerge primarily as the result of contact between two (quite dissimilar) varieties, or are they part of a dialect continuum that encompasses both Kurmanji and Sorani, a chain of dialects which differ in a series of gradual changes from each other? In the latter case, one could postulate one language “Kurdish” with the dialects Kurmanji and Sorani, which are connected by transitional dialects like Mukri. If on the other hand we consider them to result from language contact, then we would be dealing with two independent languages which converged in linguistic features due to their close contact. Mukri could represent such a contact zone or it could even be an independent variety in between Kurmanji and Sorani. The term “Kurdish” would then refer to a sprachbund (cf. Fig. 4).
Be that as it may, the varieties which share features of both areas show that there is not a clear-cut boundary between the Northern and Southern Kurdish- speaking regions. Taking up Paul’s (1998b: 171) term of “scale of northernness”, we may postulate such a scale for grammatical differences, as Blau (1989: 330) indicated for case and gender: “Les différenciations des cas et de genre disparaissent progressivement”. Whether this grammatical scale can be transferred to a geographical one requires further investigation. As a tendency, languages in the Northwest show pronominal and nominal case distinction, but lack enclitic pronouns. Languages in the Southeast make use of enclitic pronouns, but do not have other means for expressing morphological case. However, when inserting further varieties into this grammatical scale, the result is not straightforward, because some varieties, like Hawrami and Vafsi, seem to be dislocated. It remains an open question, whether this is due to migration of Sorani to the West or of Vafsi to the east and Hawrami to the South, or, whether retention or loss of case is not defined areally at all. The following table shows which varieties exhibit case for nouns (“case N”), case for free pronouns (“case Pron”), and enclitic pronouns (“encl. Pron.”). Note that Parthian is a Middle Iranian language. On this scale it is close to Persian.
The editors of this volume brought to my attention that enclitic pronouns could be lost in a language, but subsequently borrowed from a neighbouring one, while it is unlikely that gender-sensitive case would reappear after it has been lost. So the preservation of enclitic pronouns may not be a significant feature. However, at least in the case of Sorani and Southern Kurdish on the one hand, and Gorani with Hawrami on the other, borrowing of clitics can be excluded. In Old Iranian two Sandhi forms of the 3sg. existed, cf. Avestan genitive/dative -šē and -hē (i.e., *-šai and *-hai). We can be certain that these Sandhi forms were common throughout the Old Iranian languages, but in their further development, individual languages continued either one, or the other set of pronominal forms. Southern Kurdish and Sorani continued *-hai > -ī/-y, while Gorani and Hawrami continued *-šai > -š (like Persian). Thus Sorani and Southern Kurdish preserved enclitic pronouns independently from Hawrami, Gorani, and Persian.
The investigated features do not imply genetic relations, but some of them speak in favour of long contact of the respective varieties. In this paper, we also addressed the issue of chronology of grammatical changes. As mentioned in the introduction, the investigated varieties are not attested at the time when the respective changes took place. So we will compare our findings with those languages that are attested in the relevant periods. The problem with this approach is that the New Iranian languages under consideration need not have developed with the same pace, as becomes evident when we compare closely related languages such as English, Dutch, and German. Nevertheless, in the absence of direct testimonies, I use the attested Middle Iranian languages as indicators for the chronology of grammatical changes, and, as long as there are no reasons to assume a different pace, Middle Iranian findings (which themselves are not easy to pin down) will serve as benchmarks for the grammatical development (cf. Fig. 5). Needless to say, the chronology presented here is merely hypothetical.36
Since there are no traces of grammatical gender in Middle Persian and Parthian, it is certain that this grammatical category disappeared before these two languages were attested, i.e., before the 3rd c. AD in the case of Middle Persian and before the 1st c. BC in the case of Parthian. In the 3rd/4th c. AD, remnants of case are still found in the texts. In the case of Parthian, a case distinction is preserved for the free pronoun of the 1sg. throughout its attestation. Enclitic pronouns remain as well (in Middle Persian and Parthian) and are used as oblique counterparts of the free pronouns. After the 3rd c. AD, the Middle Iranian evidence increasingly suggests that a process came to an end which had already started in the Old Iranian period: the transition of the verbal system with stems marked for aspect to one with stems marked for tense. In Middle Iranian languages, specific intransitivising stem formations (most common is present -īh-, past -īhist in Middle Persian) appear. Most modern “passive” suffixes are probably related to this suffix. Sorani -r- is considered an innovation due to analogy (MacKenzie, 1961b: 84). The shift from an ergative to an accusative construction seems to have gradually proceeded after the 4th c. AD in Middle Persian, Parthian, Bactrian, and Sogdian (Jügel, 2012/I: 462). According to the Middle Persian data, it seems likely that initially, agreement of verb and object was replaced by agreement of verb and subject, before a new way of marking the logical object was grammaticalised.
When we relate the New Iranian data to the Middle Iranian observations, we can set up the following hypothetical chronology. All those varieties which preserved grammatical gender cannot be direct descendants of Middle Persian or Parthian. So we can make a distinction in Western Iranian languages, which goes back to the Early Middle Iranian period at least. In Hawrami verbal stems did not shift from aspect to tense stems. If Hawrami developed with comparable pace, then it would have become distinct around the 4th c. AD. At approximately the same time, ergative constructions began to become ambiguous so that they could be reinterpreted as accusative constructions. Hawrami, Gorani, Sorani, and Southern Kurdish would become distinct if they developed with comparable pace. However, it seems that we can refine our chronology in this respect. In a prototypical ergative construction, the logical object agrees with the verb by means of personal endings.37 In another step, topic agreement of the logical subject with coreferent enclitic pronouns evolves, which later is reinterpreted as verbal agreement. This stage seems to be represented by Hawrami. The forerunner of the corresponding construction in Sılēmani-Sorani, Gorani, and several other West Iranian varieties would have been similar to the Hawrami one. However, object-verb agreement is abandoned. The northern varieties kept the ergative construction to the present day, though dialectal tendencies towards abandoning the ergative construction can be observed in Kurmanji (cf. Dorleijn, 1996; Haig, 2008: 224ff.), generally by double-oblique constructions and the loss of object-verb agreement. The remaining features cannot be temporally localised on the basis of the Middle Iranian data. They only allow us to set up subgroups (e.g., future formation in the north). As a whole, one gets the impression that the varieties under investigation have been in close contact for a long time, and more or less positioned in the same relations on a Northwest-Southeast scale along the Zagros mountains.
A conventional tree-model assumes a common proto-language, and a series of discrete splits yielding the descendants of that common ancestor. However, in reality splits are seldom discrete, as speech communities will generally remain in contact after a split, even if they no longer constitute a single language community, and thus continue to influence each other. This is almost certainly the case for the languages under consideration here. Therefore, a combined tree model with areal effects may be more appropriate for the description of the historical developments (cf. Figures 2–4). It allows us to assume a proto language that divided, the formation of several varieties as a new language, which may later divide again, and it makes areal influence and ad-, sub-, and superstratum effects visible.
There are many more factors that influence languages and language development. The picture outlined here is just a glimpse of the complex linguistic co-relations. Nevertheless, I think that the investigation of selected grammatical features allows the constitution of areally defined linguistic units. The comparison with Middle Iranian data gives us an idea of the possible temporal localisation of Kurdish grammatical changes. Connecting such information with lexical and phonological analyses of the respective varieties, together with a history of migration, would allow us to answer questions on the linguistic unit of “Kurdish”.
References
Bedir Khan, E. Dj., and Lescot, R. (1970). Grammaire kurde (dialecte kurmandji). Paris: Libraire d’Amérique et d’Orient.
Blau, J. (1989). Le kurde. In R. Schmitt (ed.), Compendium Linguarum Iranicarum (327–335). Wiesbaden: Reichert.
Cabolov, R. L. (1997). Kurdskij Jazyk, Osnovy iranskogo jazykoznanija – novoiranskie jazyki: cevero-zapadnaja gruppa II (6–96). Moskva: Vostočnaja Literatura.
Cantera, A. (2009). On the History of the Middle Persian Nominal Inflection. In W. Sundermann et al. (eds.), Exegisti monumenta – Festschrift in Honour of Nicholas Sims-Williams (17–30). Wiesbaden: Harrassowitz.
Dorleijn, M. (1996). The Decay of Ergativity in Kurmanci. Tilburg: Tilburg University Press.
Fattah, I. K. (2000). Les dialectes kurdes méridionaux – étude linguistique et dialectologique. Louvain: Peeters.
Gippert, J. (2009). Zur dialektalen Stellung des Zazaki. Die Sprache, 47, 77–107.
Givón, Th. (1976). Topic, Pronoun and Grammatical Agreement. In Ch. N. Li (ed.), Subject and Topic (151–188). New York: Academic Press.
Haig, G. (2008). Alignment Change in Iranian Languages – A Construction Grammar Approach. Berlin: Mouton de Gruyter.
Henning, W. B. (1958). Mitteliranisch. In B. Spuler and H. Kees (eds.), Handbuch der Orientalistik. 1, Der Nahe und der Mittlere Osten, 4:1; Iranistik. 1, Linguistik (20–130). Leiden: Brill.
Huyse, Ph. (2005). Ein erneuter Deutungsversuch für den Übergang vom Schluß-y der mittelpersischen Inschriften zum Endstrich im Buchpahlavi (6.–7. Jh.). In D. Weber (ed.), Languages of Iran: Past and Present – Iranian Studies in Memoriam David Neil MacKenzie (51–67). Wiesbaden: Harrassowitz.
Jügel, T. (2009). Ergative Remnants in Sorani Kurdish? Orientalia Suecana, 58, 142–158. (Available online: http://uu.diva-portal.org/smash/get/diva2:305500/FULLTEXT02).
Jügel, T. (2012). Die Entwicklung der Ergativkonstruktion im Alt- und Mitteliranischen – eine korpusbasierte Untersuchung zu Kasus, Kongruenz und Satzbau. 2 Volumes. (Unpublished doctoral dissertation). Universität Frankfurt am Main, Frankfurt [to be published in Iranica, ed. by Maria Macuch, Wiesbaden: Harrassowitz].
Jügel, T. (2013). Die Verbalpartikel BE im Neuiranischen. Indogermanische Forschungen, 188, 299–319.
Korn, A. (2003). Balochi and the Concept of North-Western Iranian. In C. Jahani and A. Korn (eds.), The Baloch and Their Neighbours: Ethnic and Linguistic Contact in Balochistan in Historical and Modern Times (49–60). Wiesbaden: Reichert.
Korn, A. (2011). Pronouns as Verbs, Verbs as Pronouns: Demonstratives and the Copula in Iranian. In A. Korn et al. (eds.), Topics in Iranian Linguistics (53–70). Wiesbaden: Reichert.
MacKenzie, D. N. (1954). Gender in Kurdish. BSOAS, 16, 528–541.
MacKenzie, D. N. (1961a). Kurdish Dialect Studies I. London: Oxford University Press.
MacKenzie, D. N. (1961b). The Origins of Kurdish. Transactions of the Philological Society, 68–86.
MacKenzie, D. N. (1966). The Dialect of Awroman (Hawrāmān-ī luhōn). København: Munksgaard.
Mahmoudveysi, P., Bailey, D., Paul, L., and Haig, G. (2012). The Gorani language of Gawraǰū, a village of West Iran. Wiesbaden: Reichert.
Mahmoudveysi, P., and Bailey, D. (2013). The Gorani Language of Zarda, a Village of West Iran. Wiesbaden: Reichert.
Omarkhali, Kh. (2013). Kurdische Literatur. In L. Paul (ed.), Handbuch der Iranistik (340–343). Wiesbaden: Reichert.
Öpengin, E. (2013). Clitic-affix Interactions: A Corpus-Based Study of Person Marking in the Mukri Variety of Central Kurdish. (Unpublished doctoral dissertation). Université Sorbonne Nouvelle – Paris 3, Paris, and Otto-Friedrich Universität Bamberg, Bamberg.
Paul, L. (1998a). Zazaki. Wiesbaden: Reichert.
Paul, L. (1998b). The Position of Zazaki among West Iranian Languages. In N. Sims- Williams (ed.), Proceedings of the Third European Conference of Iranian Studies held in Cambridge, 11th to 15th September 1995. Part I: Old and Middle Iranian Studies (163–177). Wiesbaden: Reichert.
Paul, L. (2007). Zur Lage der Gōrānī-Dialekte im Iran und ihrer Erforschung. In M. Macuch, M. Maggi and W. Sundermann (eds.), Iranian Languages and Texts from Iran and Turan – Ronald E. Emmerick Memorial Volume (285–296). Wiesbaden: Harrassowitz.
Paul, L. (2009). Zazaki. In G. Windfuhr (ed.), The Iranian Languages (545–586). London/New York: Routledge.
Rastorgueva, V. S., and Molčanova, E. K. (1981). Parfjanskij Jazyk. In V. I. Abaev et al. (eds.), Osnovy Iranskogo Jazykoznanija (147–232). Moskau: NAUK.
Schmidt, Johannes (1872). Die Verwandtschaftsverhältnisse der indogermanischen Sprachen. Weimar: Hermann Böhlau.
Sims-Williams, N. (1996). East Iranian Languages. Encyclopædia Iranica 7, 649–652.
Skjærvø, P. O. (1997). On the Middle Persian Imperfect. In E. Pirart (ed.), Syntaxe des langues indo-iraniennes anciennes – Coloque international, Sitges Barcelona, 4.–5. Mai 1993 (161–188). Barcelona: Editorial Ausa.
Stilo, D. (2009). Case in Iranian – From Reduction and Loss to Innovation and Renewal. In A. Malchukov and A. Spencer (eds.), The Oxford Handbook of Case (700–715). Oxford: Oxford University Press.
Wendtland, A. (2011). Die Entwicklung von Demonstrativpronomen zu Artikeln im Soghdischen. Wiesbaden: Harrassowitz.
Wolff, F. (1935). Glossar zu Firdosis Schahname. Berlin: Reichsdruckerei [reprographic reprint 1965, Hildesheim: Olms].
I am grateful to Geoffrey Haig and Ergin Öpengin, whose many helpful comments improved this article significantly.
For aspects on the phonological development of Kurdish see MacKenzie (1961b) and Korn (2003).
For instance, Old Iranian nominative *čaxrah > Sogdian čaxri vs. accusative *čaxram > Sogdian čaxru vs. invariant Parthian čaxr.
For instance, Old Iranian *pāda- > Middle Persian pāy vs. Parthian/Sogdian pā
This would be along the lines of the “wave theory” (cf. Schmidt, 1872: 27).
I should point out that Old Iranian isoglosses naturally keep their relevance for later stages of the languages. However, the Middle Iranian isoglosses do not respect the Old Iranian division.
The few Bactrian inscriptions, which were known before, were rather opaque.
Due to the defective writing systems of Middle Iranian languages, we cannot be certain about the exact pronunciation.
Cf. Livšic (apud Rastorgueva and Molčanova, 1981: 188), and Huyse (2005: 52).
There are remnants of i- and u-stems in Middle Persian: -īn (< *-īnām) and -ūn (< *-ūnām). In some languages, one of the two long vowels was shortened (e.g., in Avestan -anąm, i.e., *-anām).
The ending was orthographically recycled as a marker of the end of word (Huyse, 2005).
It is unclear whether the 2nd singular had two distinct forms, i.e., tū for direct case, and tō for oblique case. The Manichaean orthography is not conclusive, and the inscriptional attestations are disputable (with ANT for tū, and L(Y)K for tō in two different text corpora).
Kurmanji ‹i› represents (approximately) the close, central, unrounded vowel, which is written ‹ı› for the other varieties.
Cf. Cantera (2009) for a historical explanation of the oblique ending.
Such a reanalysed suffix -ar for the oblique of kinship terms is found in Middle Persian and still in Early New Persian. For instance, the Šāhnāme makes use of duxt besides duxtar “daughter”, and of pus besides pusar “son” (cf. Wolff, 1935). Thus the preservation of this alternation is an archaism, which allows one to distinguish these Zazaki dialects, but it does not give any indication of when this group became distinct from its relatives.
For the Gorani dialect of Zarda, Mahmoudveysi and Bailey (2013: 30) state “Some of the pronouns can be followed by the oblique case -ī (-y following vowels). The forms of the first person singular min, second person singular to, first person plural ēma and the third person singular ād […], are all attested with the oblique suffix -ī (-y)”. I was unable to find examples in the grammatical section on pronouns (p. 31f.) or on verbal forms (pp. 43–59), nor in the glossed text 6 (pp. 79–90). However, in the section on oblique case (p. 23), one finds an indication for the forms ēma-y (1pl. oblique) and ād-ī (3sg. oblique), both governed by prepositions. Note that for the Gorani language of Gawraǰū, the authors opt to analyse a comparable “suffix”, viz. -ay, which seems to generally appear after certain prepositions, as a part of a circumposition (Mahmoudveysi et al., 2012: 12). At least, the 1pl. ēma-y could be reanalysed as ēm-ay or ēma-ay (?). Then, oblique case marking might only be attached to pronouns of the 3rd person, as it is expected.
Mahmoudveysi et al. (2012: 12, 17).
For the Mukri variety of Central Kurdish see Öpengin (2013: Chapter 2.3.2.1.1).
No such distribution is attested for the vocative in Southern Kurdish Dialects (Fattah, 2000: 257ff.). I could not find any indications on the vocative in Gorani (Mahmoudveysi et al., 2012; Mahmoudveysi and Bailey, 2013).
Article systems in Iranian languages are not unheard of, though. See, for instance, Wendtland (2011) on Sogdian.
The use of the infinitive in this construction could be a reflex of the stage where the infinitive was indifferent to diatheses as it was the case in Middle Persian (Jügel, 2012/I: Section 4.3.2).
Cf. Mahmoudveysi et al. (2012: 46) “resultative, with passive semantics” and see the explicit note in Paul (2007: 292). Gorani -īs- surely is a contraction of -īhist. This seems to be implied by Paul (l.c.) as well.
The palatal vowels may be linked to the Old Iranian derivative suffix -ya- forming intransitives.
See Bedir Khan and Lescot (1970: 163), and Haig and Öpengin (this volume).
Fattah (2000: 374 footnote 143) notes an occasionally attested calque from Arabic: râḥ ačım “je (m’en) irai”, with râḥ translated by “est allée”.
Mahmoudveysi and Bailey (2013: 43, 46ff.) claim ergative alignment for the Gorani language of Zarda. However, note that in their examples the logical object is never expressed by a full noun phrase when it is indexed by personal endings. Hence, the dialect of Zarda also seems to follow the rules set up for Sorani and the Gorani language of Gawraǰū (see in what follows).
Mahmoudveysi et al. (2012: 37) note this characteristic for logical objects only.
On reanalysis of topic agreement as verbal agreement see Givón (1976).
The agreement markers of transitive verbs are rarely attached to the verb (cf. Haig, 2008: 290ff.).
A few Iranian languages show gender distinction in the 3sg. of past tense forms.
Cf. Wendtland (2011: 86f.) for Sogdian and Pashto.
Cf. Korn (2011).
“ein areales Phänomen der jüngeren Zeit”.
The Old Iranian imperfective stem (so-called “present stem”) became the non-past stem (still called “present stem”) and the perfective stem (so-called “aorist stem”) was substituted by the Old Iranian verbal adjective in -ta- (later functioning as a past participle), which developed to the past stem.
Mahmoudveysi and Bailey (2013: 40) note the variant mi- for present tense.
It might be possible to refine the picture by taking the chronology of phonological changes into account.
If the verb still appears as a participle, agreement can also include grammatical gender and number.