Ji: Kovara Kanîzar, hejmar 4

Theodora Bynon

It is well known that during the course of their histories a number of Indo-European languages, all of them members of the Indo-Iranian branch of the family, developed an ergative construction. Thus, in certain tenses of the verb, their grammars came to treat in formally identical manner the subject of an intransitive verb and the logical object of a transitive verb, the agent (or logical subject) of this latter being given a different morphological marker.[1] Now although ergativity has been studied in a wide variety of languages as a synchronic phenomenon, the opportunities for examining it from a diachronic point of view are in the great majority of cases severely limited by lack of historical documentation. The Indo-Iranian languages, however, form a notable exception. These, thanks to the availability of texts from closely related languages covering a time-span of some three millennia, provide us with a chance to observe both the development of the construction and its subsequent decline. The present paper will do no more than attempt to trace a part of this process, namely the way in which the ergative construction has disappeared from a certain area of western Iranian. It bases its arguments upon the assumption that the geographical continuum of the Kurdish dialects, whose grammars exhibit the whole range of possibilities from fully ergative systems in the north to fully accusative systems in the south, reflects the successive stages of a diachronic process. This being granted, it should be possible by ordering the synchronic patterns of representative dialects from the northern, the central and the southern regions, to arrive at a picture of the historical sequence of events which has led to the loss of ergativity in the southern dialects and to isolate the mechanisms involved in their resultant restructuring. The wider problems of precisely how the construction arose historically and of why within the Indo-European family it appears to be confined to Indo-Iranian will not be dealt with here.[2]

The dialects of central Kurdistan can be divided at approximately the latitude of Mosul into a northern and a southern group, the line separating them following roughly the course of the Greater Zab, an eastern tributary of the river Tigris in Iraq. The northern group will be represented in our discussion by the dialects of Amadiye and Sinjar (Blau, 1975) and by a somewhat normalized variety of Kurmanji (Bedir Khan and Lescot, 1970), the southern dialects by Mukri (Mann, 1906-9) and Suleimaniye (MacKenzie, 1961). For the Suleimaniye sentences I am indebted to a native speaker of that dialect, Mr. W. O. Amin, who is at present working on a grammatical description of his language. In addition to the above, I have relied heavily for all the Iraqi dialects on the wealth of information contained in MacKenzie (1961)[3]. The transcriptions are in all cases those of my sources.

We shall first examine the relevant syntactic patterns in the northern dialects.[4] In these both the noun and the pronoun inflect for case. Apart from the vocative (which does not concern us here), there are two cases, the direct and the oblique, formally distinguished either by means of suffixes or, in the case of certain pronouns, by suppletive forms. Ergativity, as is also the case in those other Indo-Iranian languages which exhibit it, is confined to the so-called past tenses of the transitive verb. The past tenses comprise paradigms employing both simple and periphrastic constructions. The latter, which are formed by means of a participial form of the main verb and an auxiliary, will not be dealt with since they are irrelevant to the problems under discussion. The simple past tenses are derived by means of aspectual and modal prefixes from the past stem of the verb. The present tenses are derived in parallel fashion from the present stem, although they do not have an ergative construction. A simple verb form, whether present or past, consists of three elements: a prefix, a verb stem and a suffixed person-number marker (or ‘ ending ‘). The sets of person-number markers employed in the present and past tense paradigms differ formally only in the third person singular, the present tenses here having an overt marker, the past tenses zero. Thus, in Kurmanji:

Open and read as PDF file
Wek PDF veke û bixwîn

First person singular                                Third person singular


(intr.)             ez di-kev-im ‘I fall’                                   ew di-kev-e ‘he falls’

(tr.)                ez di-bin-im ‘I see’                                    ew di-bin-e ‘he sees’


(intr.)             ez ket-im ‘ I fell’                                       ew ket ‘he fell’ (tr.)

te ez dît-im ‘you saw me’ (erg.)               min ew dît ‘ I saw him’ (erg.)


Although the morphological difference is located within a single person, we will follow the practice of the grammars consulted and treat the personnumber markers as comprising two discrete sets, labelling those which go with the present stem set A and those which go with the past stem set B. The present tense is thus formally characterized by the formula prefix di + present stem + set A, the preterite by zero prefix + past stem + set B. In the examples the verbs will not be segmented, but the appropriate formula will be placed in the heading before each section so that the glosses accompanying the sentences may be confined to syntactic information. The present tense may be taken as representative of all tense and modal paradigms based on the present stem, and the preterite as representative of all those based on the past stem (pres. = present stem, past = past stem, dir. = direct case, obl. – oblique case):

Intransitive verb in the present tense (di + pres. + A):

  • hesp dikeve           ‘the horse falls’

horse-dir.                fall-3 sg.


  • hesp dikevin          ‘the horses fall’.

horse-dir.                fall-3 pl.


Intransitive verb in the preterite (zero + past + B):

  • Mirov hat                 ‘the man came’

man-dir.                  come-3 sg.


  • Mirov hatin              ‘the men came’.

man-dir.                        come-3 pl.


Transitive verb in the present tense (di + pres. + A):

  • hesp mirovî           dibîne            ‘the horse sees the man’

horse-dir.                man-obl.       see-3 sg.


  • hesp mirovan        dibîne            ‘the horse sees the men ‘.

horse-dir.                man-obl.-pl. see-3 sg.


Transitive verb in the preterite (zero + past + B):

  • s_ivên                    hesp              dît                  ‘the shepherd saw the horse’

shepherd-obl.         horse-dir.      see-3 sg.


  • s_ivên hesp              dîtin               ‘the shepherd saw the horses’

shepherd-obl.         horse-dir.      see-3 pl.


  • s_ivanan hesp              dît                  ‘the shepherds saw the horse’

shepherd-obl.-pl.    horse-dir.      see-3 sg.


  • s_ivanan       hesp              dîtin               ‘the shepherds saw the horses’.

shepherd-obl.-pl.    horse-dir.      see-3 pl.


It will be seen that in the first four sentences the noun phrase is in the direct case (the unmarked form of the noun, with zero suffix in both singular and plural) and in both the present and preterite tenses the verb agrees with it. Sentences (5) and (6) are formally parallel to sentences (1) to (4), except for the presence of a second noun phrase in the oblique case functioning as direct object. In sentences (7) to (10), however, it is the logical subject, or agent, which is in the oblique case while the logical object, or goal, is in the direct case and determines verbal concord. Mainly on historical grounds, such sentences as (7) to (10) are traditionally interpreted as passives, it being stated that in the past tenses transitive verbs obligatorily take passive form (‘das transitive Prateritum ist “passivisch”‘: Morgenstierne, 1958, 172). This analysis is, with minor reservations,[5] also the one adopted in most of the analyses of Kurdish dialects which have been consulted (MacKenzie, 1961, 193; Blau, 1975, 71 f.). There is no doubt that it fits part of the observable facts. Thus, in common with passives in general, it is the agent which is the marked form from the point of view of the morphology whereas the goal is unmarked and determines the concord of the verb so that it may be considered to be the grammatical (or’ surface ‘) subject of the sentence.[6] But, while the situation with regard to case marking and verb agreement would seem to permit the analysis of sentences (7) to (10) as passives, there are other criteria which speak against such an interpretation. Firstly, the fact that the agent is the more marked form morphologically is a phenomenon which is not limited to passives but is also characteristic of ergative constructions, in which the agent typically carries an overt morphological marker. Certain languages use a special case for this function whereas others use a case which also performs other syntactic functions, as is the situation here. Secondly, and much more importantly, sentences (7) to (10) lack a regular syntactic relationship with corresponding active sentences and thus fail to comply with a criterion which has always been considered an essential part of the definition of passive.[7]

Thus one would not normally speak of a sentence as being passive if there were not, in the language in question, a related sentence employing the same lexical items and exhibiting regular differences of syntactic structure. There would seem to be two basic views regarding the nature of the category passive. Either one may see the relationship between active and passive sentences in terms of the reversal of surface syntactic functions, the direct object of an active sentence ‘becoming’ the grammatical subject and topic (or theme) of the corresponding passive sentence with the subject of the active sentence ‘becoming’ the agent, or one may interpret passivization as the suppression of the agent, that is to say the intransitivization of a transitive verb, since it is a general characteristic of passives to require only one obligatory noun phrase, which is always the goal or logical object and never the agent. Whichever definition of the category be accepted, a passive only exists by virtue of the regular formal opposition which exists between it and a corresponding active, and of these two it is invariably the passive which is the marked category.

Now in the Kurdish dialects discussed so far, the only syntactic pattern which has a regular relationship with that illustrated by sentences (7) to (10) is the so-called agentless passive which can be regularly formed from it by deletion of the agent, conversion of the main verb to the infinitive, and the insertion of an auxiliary functioning as marker of passive voice (Bedir Khan and Lescot, 1970, 193; Blau, 1975, 72; MacKenzie, 1961, 195): e.g. Kurmanji ew hat ditîn ‘it was seen ‘.[8] Such an overt morphological marker of voice is however precisely what is absent from the verb forms of sentences (7) to (10). The verb forms of these consist of prefix + verb stem + person-number marker so that they are structurally exactly parallel to those of sentences (5) and (6). Finally, so far as the topic-comment (theme-rheme) relationship is concerned, ergative sentences have the same thematic structure as active and not passive sentences have in accusative languages in that it is the agent, and not the goal, which is the topic (or theme). By topic we mean the sentenceinitial constituent about which new information is to be conveyed, irrespective of whether or not it is also the surface grammatical subject (see Lehmann, 1976). In terms of thematic structure, therefore, the ergative construction of the past tenses of Iranian exactly parallels the accusative-type active of their present tenses (cf., for Pashto, Tegey, 1978, 24). For all these reasons we believe that sentences (7) to (10) must be interpreted as ergative and not as passive.

To summarize, in these northern Kurdish dialects the situation for the present tenses is comparable to that of any accusative language such as Latin or English, that is to say transitive sentences are structurally parallel to intransitive ones save for the presence of an additional noun phrase fulfilling the role of logical object. In the past tenses on the other hand the syntactic pattern of transitive sentences is a typically ergative one in that the noun phrase with which the verb agrees is no longer the agent but the goal and there is an additional noun phrase in left-most position performing the role of agent.[9]

If we now turn to the southern group of dialects, we find that the treatment of transitive verbs in the past tenses is not uniform throughout the area. For, while the same ergative construction with which we are already familiar is found in the northernmost of these, in the south we find a fully accusative system. Let us examine the situation in the southernmost dialect, Suleimaniye. As opposed to what we have seen for Kurmanji, neither the noun nor the pronoun of Suleimaniye inflects for case, the syntactic functions of nuclear noun phrases (that is to say, of noun phrases not preceded by prepositions) being determined by position alone. The normal or unmarked word order for transitive sentences is agent (which is also the surface subject), followed by goal (which is also the surface object), followed by verb. The verb has present and past stems and two sets (A and B) of person-number markers parallel in both phonological form and syntactic distribution to those already described for Kurmanji. Unlike Kurmanji, however, Suleimaniye also possesses a third set of exponents of person and number, quite distinct from both the personal pronouns and the markers suffixed to the verb stem. This third set of personnumber markers, which in the singular have the forms -(i)m, -(i)t, -î/y, in the plural -man, -tan, -yan, we shall label set C. Let us examine the equivalents in Suleimaniye of the ten Kurmanji sentences analysed above:

Intransitive verb in the present tense (a + pres. + A): (1′)

  • espeke ekewêt ‘the horse falls’

horse-the                fall-3 sg.


  • espekan ekewin          ‘the horses fall’.

horse-the-pl.           fall-3 pl.


Intransitive verb in the preterite (zero + past + B):

  • pyaweke hat                 ‘the man came’

man-the                  come-3 sg.


  • pyawekan hatin              ‘the men came’

man-the-pl.             come-3 pl.


Transitive verb in the present tense (a + pres. + A):

  • espeke pyaweke        ebînêt            ‘the horse sees the man’

horse-the                man-the        see-3 sg.


  • espeke pyawekan      ebînêt            ‘the horse sees the men’.

horse-the                man-the-pl.   see-3 sg.


Transitive verb in the preterite (… – C zero + past):

  • a_waneke espeke-y        bîna               ‘the shepherd saw the horse’

shepherd-the          horse-the      3 sg. see


  • s_waneke espekan-î      bînî                ‘the shepherd saw the horses’

shepherd-the          horse-the-pl. 3 sg. see


  • s_wanekan espeke-yan    bînî                ‘the shepherds saw the horse’

shepherd-the-pl.     horse-the      3 pl. see


  • s_wanekan   espakan-yan bînî                ‘the shepherds saw the horses

shepherd-the-pl.     horse-the-pl. 3 pl. see


It will be seen that in all four sentence types, irrespective of whether the verb is transitive or intransitive, present or past, the person and number of the subject are regularly repeated in a person-number marker elsewhere in the sentence. Syntactically therefore, in respect of concord, there is total systematic regularity. What is irregular, however, is the way in which concord is marked morphologically. For, while the distribution of sets A and B is the same as in Kurmanji in the case of intransitive verbs in all tenses and of transitive verbs in the present tense, in the past of transitive verbs we find as person-number markers only the members of set C and these furthermore occupy a different position from the members of sets A and B, not normally being attached to the verb stem at all but rather to the direct object. It is only when there is no overt direct object and no other suitable constituent present to which they could attach themselves that they are in fact suffixed to the verb stem (see Edmonds 1955, MacKenzie 1961, 109 for a hierarchy of potential ‘hosts ‘). Notwithstanding this irregularity, if we accept as the definition of concord the repetition of the person-number features of the subject by a morphological marker elsewhere in the sentence, we must from the point of view of its concord relations class Suleimaniye as an accusative language, and this in spite of the fact that the morphological exponents of concord employed in the transitive past are not formally or positionally parallel to those employed in the intransitive tenses and in the transitive present.

If we compare the above syntactic formulae for Suleimaniye with those given on p. 215 for Kurmanji, we see that they are identical in the case of intransitive verbs and of transitive verbs in the present tenses. For transitive verbs in the past tenses, however, not only their syntactic patterns but also their constituent structures are quite different in that, whereas in Kurmanji the marker of person-number concord (a member of set B) is a constituent of the verb, in Suleimaniye it is an immediate constituent at sentence level (a member of set C). The difference between the two dialects thus lies in the syntactic status of the person-number markers used. For it can be shown that the markers of set C in Suleimaniye are pronouns, that is to say that syntactically they have the status of noun phrases. In order to substantiate this claim, let us examine the structure of pronominal sentences in the two dialects. Examples (11) to (20) illustrate the situation for transitive verbs in Kurmanji (Bedir Khan and Lescot, 1970, 94, 176):

  • ez                 mirovî           dibînim          ‘I see the man’

I-dir. man-obl.       see/pres.-l sg.


  • ez                 mirovan        dibînim          ‘I see the men’

I-dir. man-obl.-pl. see/pres.-l sg.


  • min               mirov            kus_t             ‘I killed the man’

I-obl. man-dir.        kill/past-3 sg.


  • min               mirov            kus_tin          ‘I killed the men’

I-obl. man-dir.        kill/past-3 pl.


  • ez                 we                 dibînim          ‘I see you (pl.) ‘

I-dir. you-obl.         see/pres.-l sg.


  • min               hûn                dîtin               ‘I saw you (pl.)’

I-obl. you-dir.         see/past-2 pl.


  • hûn                                     min                dibînin           ‘you (pl.) see me’

you-dir.                   I-obl.              see/pres.-2 pl.


  • we                                      ez                  dîtim             ‘you (pl.) saw me’

you-obl.                   I-dir.              see/past-i sg.


  • ew                                      wî                  dibîne            ‘he sees him’

he-dir.                     he-obl.           see/pres.-3 sg.


  • wî                                       ew                 dît                  ‘he saw him’.

he-obl.                    he-dir.           see/past-3 sg.


It will be seen that the structure of these pronominal sentences exactly mirrors that of sentences (examples 5 to 10) in which the subject and object is a full noun phrase (that is, a phrase consisting of or having as head, a noun), this latter in each case being replaced by a personal pronoun in the appropriate case form. The equivalent sentences in Suleimaniye are:


  • min               pyaweke        ebînim           ‘I see the man’

I         man-the        see/pres.-1 sg.

  • min               pyawakan      ebînim           ‘I see the men’

I         man-the-pl.   see/pres.-1 sg.

  • min               pyaweke-m   kus_t             ‘I killed the man’

I         man-the        1 sg. kill/past


  • min               pyawekan-im                      kus_t             ‘I killed the men’

I         man-the-pl.   1 sg. kill/past


  • A:
    min    êwe               ebînim           ‘I see you (pl.) ‘
    I         you-pl            see/pres-1 sg.


  • A: min êwe-m bînî ‘I saw you (pl.)

B: bînî-m-in

  • A: êwe min ebînin                        ‘you (pl.) see me’
    B: em-bîn-in


  • A: êwe min-tan bînî                                           ‘you (pl. saw me’

B: bînî-tan-im


  • A: ew ew ebînêt                           ‘he sees him’

B: ey-bîn-êt


  • A: ew ewî bînî
    B: bînî-y


In these Suleimaniye examples it will be noted that, when both subject and object are pronominal (sentences 15′ to 20′), there are two possible ways of expressing the same meaning, one (15’a to 20’a) using personal pronouns, the other (15’b to 20’b) relying solely on the person-number markers. In the first of these constructions the pronouns behave like full noun phrases, and there is no problem. In the second it will be seen that all three person-number marker sets (A, B and C) are used but with a well-defined distribution: set A has logical subject function throughout and is confined to the present tenses, set B has logical object function throughout and is confined to the past tenses, set C functions as both logical subject and logical object dependent upon tense.

Looking at this thoroughly complex pattern in terms of morph sequence, it will be seen that the members of set C generally precede those of sets A and B, and furthermore that they immediately follow the initial morph of the verbal complex irrespective of its status (prefix in the present, verb stem in the past). Their behaviour in these verbal complexes in fact exactly parallels their behaviour in sentences with full noun phrases (sentences 7′ to 10′) or personal pronouns (sentences 18’a, 20’a), where they regularly attach themselves to the first constituent other than the logical subject. These positional properties of the members of set C, taken in conjunction with the fact that they never occur initially in a sentence or phrase, point to their clitic nature. Their syntactic functions on the other hand inescapably identify them as pronouns for, apart from acting as direct and indirect object of a transitive verb in the present tenses and as agent in the past tenses, they also occur in ‘genitive’ relation to a noun phrase (kitêb-yan ‘their book’) and may be governed by a preposition (lê-man-î sendin ‘he took them from us ‘: MacKenzie, 1961, 114).

961, 114). While the pronominal status of set C has never really been questioned, the fact that (in spite of their formal and distributional parallelism with set B of Kurmanji) the members of set B in Suleimaniye are also pronouns is less obvious-although it was clearly stated by Edmonds as early as 1955.[10] For whereas in the intransitive past the members of set B in Suleimaniye might conceivably be interpreted as the ‘personal endings’ of the verb, such an analysis fails to take account of the fact that in the transitive past (which, unlike the situation in Kurmanji, is not of course ergative) they fulfil exactly the same functions as are performed in the present tenses by set C, namely those of direct object and indirect object (noun phrase governed by preposition). With respect to this latter function, compare the following sentences in which the preposition bo ‘to, for’ occurs first with a present and then with a past tense verb form (Edmonds, 1955, 498, 502):

diyariy-êk-tan                                           bo                  e-hên-în        ‘we shall bring a present for you’
present-one-2 pl.                for                 pref.-bring/pres.-1 pl.

in which bo governs tan (2 pl. C)


Xwa               bo-y               nard-im-î(t) ‘                                            God sent you to me’
God               for-3 sg.        send/past – 1 sg. B – 2 sg.

in which bo governs im (1 sg. B). The choice between sets B and C is, it will be seen, dependent solely upon the tense of the verb. Only the ‘genitive’ construction may employ set C in all tenses (see MacKenzie, 1961: 229)

We have so far considered the evidence for the pronominal status of sets B and C. Can the members of set A also be classed as pronouns ? We have seen that their syntactic distribution is much more limited than that of sets B and C, their only function being to mark agreement of the subject with the verb in the present tenses. We have however also noted (p. 212) that the only formal difference between its members and those of set B lies in the third person singular, all the other forms being identical. There is furthermore no environment in which these two third person forms can contrast, the A form occurring after the present stem of the verb, the B form after the past stem or after a member of set C. There might thus be grounds for treating the third person suffixes as allomorphs and collapsing sets A and B into a single paradigm so that all person-number markers will be pronominal. Such a choice would have important consequences for the analysis of verbal structure in Suleimaniye, for, if it is correct and we are to consider all person-number markers as pronouns, this necessarily implies that Suleimaniye does not have the ‘personal’ verb of the majority of Indo-European languages and that instead we have an ‘impersonal’ verb whose pronominal satellites do not signal the subject in any way differently from the direct and indirect objects.[11]

Alternatively, instead of basing ourselves upon the formal similarities of the markers, we may make the basis of our analysis the distributional differences between the three sets, although this entails in the case of set B operating with homophonous paradigms. MacKenzie thus interprets set A as ‘purely verbal’ (that is to say, as serving solely as verbal endings), set B as both verbal and pronominal (that is, as having two formally identical subsets, one being verbal endings the other pronouns), and set C as pronominal only. If, however, we adopt this latter view, thus retaining the concept of a personal verb, we must bring the verb forms of the transitive past into paradigmatic alignment with those of the transitive present and of the intransitive verb, both of which agree with the subject in person and number, and interpret the absence of a personal ending in the transitive past as an instance of zero marking. The resultant zero suffix which has to be posited for each person in the transitive paradigm will then be an allomorph of the corresponding member of set B of the intransitive past and like it should logically signal agreement with the subject. In MacKenzie’s analysis, however, it is taken as a marker of agreement with the logical object. Although there can be no doubt that this analysis is historically correct and accords with the synchronic structure of related dialects including Kurmanji, it cannot be justified on purely internal and synchronic grounds. For, as we have seen, nowhere else in Suleimaniye is there overt agreement of the verb with the logical object.

We have seen that if we opt for zero morphs in the transitive past we must treat these as allomorphs of the overt markers in the intransitive past, signalling agreement with the subject. And indeed in Persian, a language which is both contiguous and closely related historically, the endings of the intransitive past were analogically transferred to the transitive past-a regularization process which clearly presupposes an identical synchronic analysis. This solution has however at least one obvious drawback in that it disregards the possibility of an overt marker in the form of the clitic pronoun in favour of a covert marker as the locus of concord. Indeed, on present evidence at least, none of the synchronic analyses so far proposed would appear capable of relating the morphological structures to the syntactic rules in a simple and straightforward manner.

Although the complexity of the relationship which exists between the morphology and the syntax of Suleimaniye may appear strange from a synchronic point of view, its explanation in terms of the history of the language is obvious enough. For it is clear that, while restructuring its syntax in an accusative direction, it has retained the bulk of the morphology of an earlier ergative state. The manner in which this syntactic restructuring is most likely to have come about can be seen if we compare the verbal morphology of Suleimaniye with that of Kurmanji. In our descriptions of the two dialects we equated their sets A and B of person-number markers on the grounds of shared phonological form and distribution in verbal constructions. Classing them on these same grounds as ‘ diachronic sames ‘, we can see that the first important difference between the northern dialects and Suleimaniye is the absence of the ‘ ergative ‘ verbal endings from the transitive past in the latter. It must have been the loss of these endings in Suleimaniye which brought about the syntactic restructuring. For with their disappearance there no longer remained any overt link between the verb and the object so that the object lost its status as surface subject and this property now passed to the agent. We must presumably seek the starting-point for this structural change in the most common sentence-type, in which the logical object of the (still ergative) construction is either a full noun phrase in the singular or a third person singular pronoun, so that the verbal ending will in either case be zero (see examples 13 and 20 above). In such a sentence there is thus no overt marker of agreement of the verb with either the logical subject or the logical object. We must assume that the suffixless verb form spread from this sentence-type to sentences with first or second person object in the singular and with object in any person in the plural, resulting in the loss of all formal link between verb and logical object.

Two ways in which the status of surface subject was transferred from the logical object to the agent may be envisaged. If we accept the synchronic analysis which postulates zero endings for the transitive past we may simply assume that after the loss of the old ‘ ergative’ endings their absence was interpreted as zero exponency of concord and, since the only overt markers of concord elsewhere in the language all referred to the agent, the transitive past was reanalysed accordingly. This solution it will be noted treats the change from ergative to accusative concord in isolation, without relating it in any way to the second major development which took place in Suleimaniye, namely the insertion of an obligatory pronominal copy of the agent in past-tense transitive sentences. It would, however, appear equally feasible that this pronominal copy of the agent was instrumental in bringing about the syntactic restructuring since, following our alternative synchronic analysis, the clitic pronoun would, in the absence of an overt verbal ending, have had to be reinterpreted as marking concord with the agent. One good argument in favour of this interpretation is the fact that sentences without the clitic pronoun are totally lacking in Suleimaniye.

Before attempting to decide in favour of one solution rather than the other, let us then examine this second major development in greater detail. We have seen that Suleimaniye differs from Kurmanji in possessing the personnumber markers of set C, which are absent from Kurmanji. Now we know that the members of set C continue in both phonological form and syntactic distribution the Old Iranian clitic pronouns a characteristic of whose paradigm was the absence of a nominative form. By Middle Persian, the closest relative to the ancestor of the modern Kurdish dialects for which we have written data, their case paradigm had been reduced to a single form per person, they were suffixed to the first suitable item in the sentence and they had the same syntactic function as a noun or pronoun in the oblique case (Henning, 1933, 242 f.). This latter was employed, as in the northern dialects of Kurdish, to mark the object of a transitive verb in the present and the agent in the ergatively constructed past. Given the fact that the ancestors of set C were clitic pronouns having the distribution of an oblique case, its syntactic distribution in Suleimaniye is fully accounted for with the sole exception of its use in the transitive past, where it obligatorily repeats the syntactic features of a nominal or pronominal agent. For we have seen that in both Middle Persian and northern Kurdish a noun phrase in the oblique case could by itself fulfil the role of agent. Suleimaniye, therefore, would appear to have innovated by obligatorily inserting a clitic pronoun repeating the person and number of the agent. Now an exactly parallel formal construction is still used in Suleimaniye for the purpose of topicalizing any nominal constituent other than the agent. If such a nominal constituent is made the topic, it is moved to the head of the sentence and its person and number features are repeated by affixing the appropriate member of set C to the constituent following the agent. Thus the first of the following two sentences is thematically unmarked whereas in the second the direct object has been topicalized[12]:

min                xwardinakan-im                 xward                                  ‘I ate the food’
I                     foodstuff-the-pl.-1 sg.         eat/past

xwardinakan                                             min                xward-im-in ‘as for the food, I ate it ‘.
foodstuff-the-pl.                 I                     eat/past-I sg. C-3 pl. B


This suggests that the syntactic pattern characterized by the pronominal copy of the agent was originally also a marked one and that topicalization of the agent must have been generalized to become the unmarked norm.

Further evidence regarding the nature of the topicalization rule and its place in relation to other syntactic constructions may be adduced from dialects of the southern group which occupy an intermediate position between the two extremes so far described. Thus Mukri has both the ergative construction and two case system of Kurmanji and the northern group and the construction with pronominal copy of the agent of Suleimaniye. Mukri has in fact two syntactic possibilities for rendering the transitive past. The rarer of these two constructions is that of the northern dialects, which places the agent in the oblique case and lacks the pronominal copy. The more usual one is that found in Suleimaniye with the pronominal copy of the agent, although it should be noted that here the agent is in the direct and not the oblique case. Mann (1906, lxxxv f.) accounts for this latter pattern in the following way. Starting from the Old Iranian passive, from which the ergative construction arose,13 he argues that a sentence such as
seg                hel-î-girt                              ‘the dog took it (the ring)’
dog-dir.         pref.-3 sg. -take/past-zero


is to be understood as having originally meant ‘the dog, it (the ring) was taken by him ‘, that is with both noun phrases in the nominative (the ancestor of the direct) case, but for different reasons. It is the appropriate case for the first noun phrase because this is placed ahead of the sentence proper as the topic or theme, and it is the appropriate case for the second because it is the reflex of the grammatical subject of an earlier passive. Note that this analysis, which assumes that the topicalized agent precedes the sentence proper and is thus not a constituent of it, neatly explains the position of C in the sentence: being clitic, it cannot occur sentence or phrase initially and the earliest possible slot it can occupy is immediately after the first constituent.

We are now in a position to attempt a synthesis of the historical developments which must have taken place in the various dialects. In the northern group the most important innovation was the loss of the clitic pronouns (see MacKenzie, 1961, 222) and with them the loss of the topicalization rule. The southern group, while retaining these pronouns together with the topicalization rule, innovated to varying degrees. Of all the Kurdish dialects those intermediate between the northern group on the one hand and Suleimaniye on the other may perhaps be considered the most conservative. Thus the only significant development in Mukri seems to have been the almost total replacement of the simple (ergative) construction of the transitive past by the thematically marked (though still ergative) construction with topicalized agent. In Suleimaniye this replacement process has reached completion, for we have seen that it lacks all trace of the original unmarked construction and has made topicalization of the agent obligatory so that this latter construction has ceased to be a marked one. The complete loss of the simple construction lacking a pronominal copy of the agent may in fact constitute an argument in favour of treating the clitic pronoun, and not the verb, as the locus of concord in the transitive past. For, due to the loss of case, the expected reflex of the unmarked construction in Suleimaniye would have been a sequence of two noun phrases without any case marking followed by a verb without any personnumber marker. To judge from results, this construction does not appear to have been a viable one and simply did not survive. Apart from losing the unmarked construction of the transitive past and case as a morphological category, the most important innovation of Suleimaniye has been the loss of the ergative construction.[13] As we have seen, this was brought about by the loss of the ‘ ergative ‘ endings from the verb in the transitive past, leading to the conflation of agent and surface subject. Since Suleimaniye has at the same time retained a good deal of the verbal morphology of an earlier state in which the syntax was still ergative, the restructuring at the syntactic level has resulted in an extremely complex relationship between morphology and syntax.

Finally the identification of the surface subject with the agent must also have been responsible for the reinterpretation as pronouns, in constructions such as (16’b), (18’b), and (20’b), of the person-number markers of set B. For, despite the fact that these latter are historically the endings of the verb marking ergative concord, the reinterpretation of the members of set C as obligatory copies of the agent must necessarily have entailed the members of set B losing their surface subject status, so that the only role they have retained is that of representing the object, now both logical and surface. As a result, the old verbal endings have (in this construction at least) become pronouns, although they have kept their position as appendices of the verb, thereby increasing the discrepancy between morphological form and syntactic function.



  • Allen, W. S. 1950-1. ‘A study in the analysis of Hindi sentence-structure ‘, Acta Linguistica, vi, 68-86.
  • Allen, W. S. 1964. ‘ Transitivity and possession ‘, Language, XL, 337-43. Anderson, S. R. 1976. ‘ On the notion of subject in ergative languages’ in C. N. Li (ed.), Subject and topic. New York: Academic Press, 1-23.
  • Bedir Khan, Emir Djeladet and Lescot, R. 1970. Grammaire kurde (dialecte kurmandji) (Librairie d’Amérique et d’Orient). Paris: Adrien Maisonneuve.
  • Benveniste, E. 1952. ‘ La construction passive du parfait transitif ‘, BSLP XLVIII 52-62.
  • Blau, Joyce. 1975. Le kurde de ‘Amadiya et de Djabal Sindjar. Paris: Librairie C. Klincksieck.
  • Cardona G. 1970. ‘ The Indo-Iranian construction mana (mama) krtam’ Language XLVI, 1-12.
  • Comrie, B. 1976. Ergativity. Typescript.
  • Edmonds, C. J. 1955. ‘ Prepositions and personal affixes in southern Kurdish ‘, BSOAS, xvn, 3, 490-502.
  • Henning, W. 1933. ‘ Das Verbum des Mittelpersischen der Turfanfragmente ‘, ZII, ix, 158-253. Lehmann, W. P. 1976. ‘ From topic to subject in Indo-European ‘ in C. N. Li (ed.), Subject and topic. New York: Academic Press, 445-6.
  • MacKenzie, D. N. 1961. Kurdish dialect studies, I. London: Oxford University Press.
  • Mann, 0. 1906-9. Die Mundart der Mukri-Kurden. Berlin: Georg Reimer.
  • Morgenstierne, G. 1958. ‘Neu-iranische Sprachen’ in Handbuch der Orientalistik, 1. Abt., Band iv, 1. Abschnitt. Leiden: E. J. Brill, 155-221.
  • Schmidt, K. H. 1963. ‘ Zum Agens beim Passiv ‘, IF, Lxvir, 1-12. Schmidt, K. H. 1973. ‘ Transitiv und Intransitiv ‘ in G. Redard (ed.), Indogermanische und allgemeine Sprachwissenschaft: Akten der IV. Fachtagung der Indogermanischen Gesellschaft. Wiesbaden: Dr. Ludwig Reichert Verlag, 107-24.
  • Schmidt, K. H. 1977. ‘ Probleme der Ergativkonstruktion ‘, Manchner Studien zur Sprachwissenschaft, xxxvi, 97-116.
  • Tegey, H. 1978. ‘Ergativity in Pashto ‘, Pashto Quarterly, I, 3, 3-88.
  • Wagner, H. 1978. ‘ The typological background of the ergative construction ‘, Proceedings of the Royal Irish Academy, LXXVm , C, 3, 37-74.









[1] In Basque, a typical ergative language, the subject of an intransitive verb is in the same ‘absolute’ case as the direct object of a transitive verb and the subject of this latter is in the ‘ ergative ‘ case. Thus gizon ‘ man ‘ is in the absolute case (marked by zero suffix) in both gizon ethorri d-a ‘ a man has come’ and aita-k gizon ikusi d-u ‘the father has seen a man ‘, aita-k ‘father’ in this latter being in the ergative, or ‘ agentive’, case (marked by the suffix -k). In addition, the prefix d- of the final auxiliary verb in both sentences refers to gizon so that one could speak of the verb as ‘ agreeing ‘ with gizon were it not for the complicating factor that in addition to this prefix the auxiliary also carries a personal ending (-t in the first, -k in the second, zero in the third person) which marks the logical subject. Thus in aita ikusi d-u-t ‘I have seen the father ‘ the prefix d- reflects the logical direct object in the absolute case aita and the suffix -t marks the logical subject ‘ I’: see Wagner (1978), 38. Similarly in Georgian, iremi ‘ stag ‘ is morphologically identical in iremi igo tgeAi ‘the stag was in the wood’ and monadire-m mokla iremi ‘ the hunter killed the stag ‘, the agent or logical subject in the latter sentence again having a distinctive marker -m: see Schmidt (1973), 109. For list of references see p. 224.

[2] For a discussion of the areal, typological and historical aspects of ergativity on a world scale and a comprehensive bibliography, see Wagner (1978); for ergativity in Indo-Iranian see e.g. Schmidt (1973), 116, notes 36, 37; for Hindi see Allen (1950); for hypotheses regarding ergativity in Proto-Indo-European, see Schmidt (1977).

[3] I am also grateful to Professor D. N. MacKenzie for valuable criticisms and comments on an earlier version of this paper.

[4] Bedir Khan and Lescot (1970), 176 f., 93 f.; cf. Blau (1975), 48, 71; MacKenzie (1961), 106 f., 155 f., 193 f.

[5] MacKenzie (1961), 193 makes the point that the past tenses of transitive verbs are conjugated in the same manner as those of intransitive verbs.

[6] Thus, in the active sentences puer puellam amat: the boy loves the girl, the agent is in the least marked form whereas in the corresponding passive sentences it is much more marked: puella a puero amatur: the girl is loved by the boy. The claimed isomorphism of passive and ergative constructions (Schmidt (1973), 111) is, however, only partial even in this lespect, since there are often far more cogent syntactic arguments for the ‘ subjecthood ‘ of the surface subject of a passive than for that of an ergative construction; see Anderson (1976) for various syntactic tests for subjecthood. We will in this paper adhere to the traditional view that the noun phrase which determines verbal concord is the surface subject.

[7] F7 or a brief survey of opinions, from Brugmann to Chomsky, see Schmidt (1963), 2 ff.

[8] In Kurmanji, as in a number of other languages, ergative and (agentless) passive constructions coexist and there is thus no systematic incompatibility of the two as has been claimed (see Schmidt (1973), 111).

[9] This analysis entails that in minimal sentences with a transitive verb and a single noun phrase this noun phrase will be interpreted as agent in the present but as goal in the past. This point could not be tested since the sources consulted lack the relevant information. It would, however, seem likely that neither of the two noun phrases of a transitive sentence may be absent.

[10] The statement is worth quoting in full: ‘The pronouns in their separable forms may be used to perform any function of a noun [read: noun phrase] but, unless emphasis is intended, it is generally more idiomatic to use the affixes [that is to say our sets B and C] ‘: Edmonds (1955),,491.

[11] Thus Wagner (1978), 46 ff. considers the non-orientation of the verb to be a typical feature of ergative languages; although now no longer ergative in fact, Suleimaniye would on this basis qualify as an ergative language.

[12] For a return to this older interpretation, despite Benveniste (1952), see Cardona (1970).

[13] N. J. Sims-Williams has suggested that the loss of the ergative construction in Kurdish is likely to be closely linked with the loss of the two-case system. I am grateful to him for this and other comments as well as for drawing my attention to the papers by Comrie and Tegey.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s