ERGATIVITY in SULEIMANIYE KURDISH

Robyn C. Friend, Ph.D.

Originally presented at the Middle East Studies Association 1985 Annual Meeting, November 1985

Rights statement:  This material is the intellectual property of the author.  We encourage hyperlinks to be made directly to this URL.  We encourage recognized academic usage, including quoting (with appropriate credits) and citing.  We also encourage educational use, and publication in commercial forums; permission for such uses can be obtained by writing to:

Snark Records
c/o Robyn Friend
19 Golden Spar Place
Rolling Hills Estates, California  90274

Such use is authorized only after obtaining written permission in advance from Robyn C. Friend.  Thank you.


Abstract:  Suleimaniye Kurdish, an Iranian language spoken in Suleimaniye, Iraq, shows evidence of both ergative/absolutive and nominative/accusative structures.  Using test criteria from the field of relational grammar, it is possible to determine the presence or absence of ergativity in the syntax and morphology of Suleimaniye Kurdish.  The results of the application of these tests show that while there is no solid evidence of ergativity in Suleimaniye Kurdish syntax, there is evidence of ergativity in the morphology, more specifically, in the past-tense transitive construction.

1.  INTRODUCTION

In this paper, I will discuss ergativity specifically with regard to Suleimaniye Kurdish.

In the case of any one particular language, it may not always be obvious whether the grammar is ergative/absolutive or nominative/accusative; this is true partially because grammars which employ features of the ergative/absolutive system tend to employ a combination of both systems, rather than being purely ergative.  Indeed, Moravscik claims that “all human languages have both some ergative and also some accusative patterns” [ 1978a:275 ].

In this paper I will define a set of criteria which attempts to permit the following determinations:  (a) whether or not a particular language exhibits characteristics of ergativity in certain aspects of its morphology; and (b) whether or not a particular language exhibits characteristics of ergativity in certain aspects of its syntax.  Note, however,  that such tests should be considered indicative rather than conclusive; not all these tests have an equal chance of giving results, and even the most thoroughly ergative language may show some nominative/accusative structures in its grammar (e.g., pronouns in Dyirbal show nominative/accusative inflection [ Dixon 1979:64 ] ).  The theoretical bases for these criteria are drawn in part from work by linguists in the fields of relational grammar and language universals; specific authors are cited in the text.  The prime innovators of such tests and their use in this fashion are S. Anderson [ 1976 ] and R.M.W. Dixon [ 1979 ].  Once I have established these criteria, I will  apply them to the grammar of Suleimaniye Kurdish.

2.  EXAMINATION of SULEIMANIYE KURDISH for MORPHOLOGICAL ERGATIVITY

Ergativity in the morphology of a language manifests itself through the means employed for marking (identifying) grammatical function [ Anderson 1976:3 ].    Means which mark subjects and objects alike and agents differently might be said to show ergative marking [ see examples 1 and 2 ].  Keenan [ 1976:324-5 ] identifies three coding properties of noun phrases which may be examined for evidence of ergative morphology. These coding properties are position (i.e., constituent order), case marking, and verb agreement.

2.1  Position

Position refers to the position or order of constituents within a sentence.  If the order of constituents, especially the relative position of noun phrases with regard to the verb, in a transitive sentence differs from that of an intransitive sentence, this might be considered evidence of ergativity.  For example, in ergative constituent order, subjects and objects would have the same position in relation to the verb, and agents would have a different position.  An ordering of AVO (agent-verb-object, regardless of the presence of other constituents) for transitive verbs, and VS (verb-subject) for intransitive verbs would indicate ergativity; an ordering of OVA for transitive verbs, and SV for intransitive verbs, also would indicate ergativity [ Dixon 1979:67 ].

The test of constituent position is only truly useful when applied to verb-medial languages. Because Kurdish is a verb-final language, this test does not yield any relevant evidence in support of the presence of ergativity in Suleimaniye Kurdish grammar (verb-medial word-order is possible in Suleimaniye Kurdish, though only in highly-marked circumstances, and therefore not useful for this purpose).

2.2  Case Marking

Three types of case marking are considered: case endings, markedness of pronouns, and markedness of clitics.

Case Endings:  In an ergative/absolutive system, subjects and objects are marked with the same case (usually the unmarked, or absolutive, case), and agents are marked with a different case (the marked, or ergative, case) [ Dixon 1979:62, 64 ].  However, Suleimaniye Kurdish displays case marking neither in the present tense nor in the past tense.  Therefore, a study of case endings in Suleimaniye Kurdish will reveal no information concerning the presence or absence of ergativity.

Markedness of Pronouns:  Parallel to case-marking is the marking of pronouns.  In an ergative/absolutive system there may be both marked and unmarked sets of pronouns, as in the Tati language group and other Iranian languages [  Stilo 1981:147 ]. In such a system, subject and object pronouns are unmarked, while agent pronouns are marked.

The data from Suleimaniye Kurdish include no indications of the existence of two classes of pronouns in the present tense.  Note in examples 3a and 3b from MacKenzie [ 1962 ] that /min/ may be a subject in one sentence, and an object in another, and an agent in another.  There do not seem to be two classes of pronouns in Suleimaniye Kurdish in the past tenses, either.  In examples 3d and 3e, the pronoun /aw/ ‘he’, appears in the first example as an agent, but in the second example as a direct object.

Therefore, the data on pronoun marking yield no information on the presence or absence of ergativity in Suleimaniye Kurdish.

Markedness of Clitics:  A language may employ clitics, constituents which have no independent existence, but rather are attached to another constituent in the sentence [Comrie 1981:21].  These clitics can act as bound pronouns (on the verb, and on other consitutents) which record in the sentence the grammatical relations subject-of, object-of, or agent-of.  If the set of clitics which represent subject and object are different from the set of clitics which represent agent, this might be evidence of ergative morphology [ Dixon 1979:66 ].

Suleimaniye Kurdish has two sets of clitics, which I refer to as Set 1 (unmarked), and Set 2 (marked) [ after Stilo 1981 ].  In the present tense, the subject and agent are represented in the clitic form by the Set 1 (unmarked) clitics, and the object is represented by the Set 2, marked clitics, as in the following diagram:
 
  Clitic Marking, Present Tense:

        A   DO
        1   2

          S
          1

The subject ‘we’ in example 4a and the agent ‘she in example 4b are both represented by Set 1 clitics, whereas the object ‘me’ of 4b is represented by a Set 2 clitic.

In the past tenses, however, it is the subject and object which are represented in the clitic form by Set 1 clitics, and the agent is represented by a Set 2 clitic, as illustrated in the following diagram:

  Clitic Marking, Past Tenses:

           A  DO
           2  1

            S
            1

In examples 4c and 4d, the subject of 4c and the object of 4d are represented by Set 1 clitics, and the agent of 4d is represented by a Set 2 clitic.  Thus, in the past tense, subject and object are treated as a class, apart from agent, which is treated separately, while in the present tense, subject and agent are treated as a class, and object is treated separately.  It would appear that with regard to the markedness of clitics in Suleimaniye Kurdish, clitic morphology is ergative in the past tenses, but not in the present tense.

2.3  Verb Agreement

Another morphological feature which may have implications for determining the degree of ergativity in a language is verb agreement.  If the verb shows agreement with the subject of an intransitive verb and with the direct object in one way, and with agent in another way, this may be evidence of ergative verb agreement [ Trask 1979:388 ; Dixon 1979:66 ].
 
Determining exactly which features of the morphology of a language constitute verb agreement is not straightforward.  Moravscik [ 1978b:333 ] defines the broader notion of agreement as follows:

 “… a grammatical constituent A will be said to agree with a grammatical constituent B in properties C in language L if C is a set of meaning-related properties of A and there is a covariance relationship between C and some phonological properties of a constituent B1 across some subset of the sentences of language L, where constituent B1 is adjacent to constituent B and the only meaning-related non-categorial properties of constituent B1 are the properties C”.

To re-phrase this statement to represent notions specifically about verb agreement, one might say the following:  agreement is shown between a noun phrase (A) and a verb (B) in person and number (C) (or any other such meaning-related properties) if person and number (C) is a set of meaning-related properties of the noun phrase (A), and there is some covariance relationship between person and number (C) and some phonological properties of the agreement clitic (B1) across a subset of sentences in the language, where the agreement clitic (B1) is adjacent to the verb (B) and the only meaning-related non-categorial properties of the agreement clitic (B1) are the properties of person and number (C).

This definition of agreement presents some obvious problems for describing the system of verb agreement in Suleimaniye Kurdish.  First, verb agreement clitics in Suleimaniye Kurdish are not necessarily adjacent to the verb.  For example, there are cases where the clitic which shows agreement with an agent moves off of the verb and attaches to an earlier constituent.  Second,  the agreement clitic may not have an overt referent, i.e., there may not be an expressed noun phrase to which the clitic refers, thus showing agreement between the noun phrase and the verb; the clitic may be the only reference to an otherwise unexpressed agent, subject, or object.  Third, there may be more than one clitic in the sentence, and both may be on the verb or elsewhere.  How then to determine what constitutes verb agreement in Suleimaniye Kurdish?

For the purposes of describing verb agreement in Suleimaniye Kurdish, I must go beyond Moravscik’s definition and include as verb agreement any kind of pronominal morphology which refers to major noun phrases of the predicate.  Clitics on the verb are examples of such pronominal morphology, but verb agreement clitics may not necessarily be bound to the verb.

As I see it, there are three main criteria for determining verb agreement in Suleimaniye Kurdish:  (1) clitic-type (i.e., Set 1 or Set 2); (2) whether the clitic is attached to the verb; and (3) circumstances under which a clitic is required.

With regard to criterion (1), clitic-type, one might consider that verb agreement is expressed solely by Set 1 clitics.  The rationale for this would be the following: the relationship between the verb and its agreed-with noun phrase is an unmarked relationship, and therefore should be indicated by unmarked Set 1 clitics rather than marked Set 2 clitics.  For example, in Modern Persian, verb agreement is indicated by Set 1 clitics in all tenses, regardless of the transitivity of the verb.  In fact, in Suleimaniye Kurdish, Set 1 clitics indicate verb agreement in three out of the four tense/transitivity possibilities.  In the present tense, regardless of the transitivity of the verb, a Set 1 clitic on the verb indicates agreement with either the subject or the agent.  A Set 2 clitic on the verb in the present tense indicates the implied presence of a direct object.  In the past tense, agreement is split along lines of transitivity:  in a past-tense intransitive sentence, a Set 1 clitic on the verb indicates agreement with the subject; however, in a past-tense transitive sentence, a Set 1 clitic on the verb refers to the direct object of the verb, and not to the agent.  The clitic which refers to the agent is Set 2, and may not be attached to the verb at all.  Thus criterion (1), clitic-type, shows nominative/accusative agreement in the present tense, where Set 1 verb agreement clitics refer to subject and agent, contrasting with ergative agreement in the past tenses, where Set 1 clitics refer to subject and object, as we have seen in the previous discussion and in examples 4a, b, c, and d.

One might also consider that the criteria for determining verb agreement in Suleimaniye Kurdish is based on criterion (2), the clitic which is attached to the verb.  This approach presents problems in that an object clitic (both direct and indirect) may appear on the verb, while the fronting of the agent clitic in the past tense, may leave only an object clitic, or no clitic at all, on the verb.  As examples 4a, b, c, and d show, in the present tense, subject and agent clitics are suffixed to the verb, while direct object clitics may also appear on the verb, before the verb stem.  In the past tenses, subject and object clitics are suffixed to the verb, but agent clitics move off the verb.  Thus criterion (2) yields mixed results: nominative/accusative patterns in the present tense, and ergative/absolutive patterns in the past tenses.

A final criterion for consideration might be criterion (3), whether or not a clitic is required at all.  In both present and past tenses in Suleimaniye Kurdish, agent and subject clitics co-occur with overt noun phrases, but object clitics never do.  That is, in the case of subject and agent, a clitic is required to appear on the verb (and then move off, in the case of past transitive verbs) regardless of whether or not an explicit subject or agent noun phrase is present in the sentence.  However, an object clitic appears only when there is no explicit object present in the sentence.  In other words, object clitics act only as bound pronouns, bound either to the verb or to some other eligible constituent (e.g., with independent adpositions), and refer only to an implied object, and not to an expressed one.   Clitics are required for both subject and agent, regardless of tense, and regardless of whether or not an explicit noun phrase referent is present in the sentence.  Clitics are required for direct objects only if there is no explicit noun phrase referent present, regardless of tense.  Thus criterion (3) shows nominative/accusative agreement in all tenses and transitivities.

Summary of Verb Agreement:  If we review these three criteria, we find that verb agreement in Suleimaniye Kurdish operates in a split manner, that is, the verb agreement system contains elements of both ergative and nominative/accusative marking.  Ergative verb agreement appears in the past tenses in the choice of clitic type and in the choice of which clitic may appear on the verb; nominative/accusative agreement appears in the present tense in the choice of clitic type and in the choice of which clitic may appear on the verb, and in all tenses in the obligatoriness of subject and agent agreement clitics.

2.4  Summary of Morphological Criteria

The examination of the morphology of Suleimaniye Kurdish for ergativity has yielded mixed results.  Since constituent order in Suleimaniye Kurdish is not verb-medial, there is no definitive contrasting constituent order between transitive and intransitive sentences.  Since neither case endings nor marked-versus-unmarked sets of pronouns exist in Suleimaniye Kurdish, these  criteria yield no information regarding ergative/absolutive versus nominative/accusative morphology in Suleimaniye Kurdish.

However, some evidence of ergative morphology appears in the choice of clitics used for each grammatical relation, and to some extent, in verb agreement.  In both of these two aspects of Suleimaniye Kurdish morphology, whatever evidence there is for ergativity appears only in the past tenses, with present-tense marking and verb agreement proceeding along nominative/accusative lines.

3.  EXAMINATION of SULEIMANIYE KURDISH for SYNTACTIC ERGATIVITY

The measure of syntactic ergativity present in a language can be gauged by the extent to which the application of syntactic rules differentiates subject and object as a class separate from agent.  The theory of relational grammar, in its narrowest sense, holds that the grammatical relations of subject-of, direct-object-of and indirect-object-of are language primitives, to which the rules of syntax apply [ Perlmutter and Postal, 1983 ].  In this general spirit, I consider grammatical relations to be primitives, but, like Dixon [ 1979 ] (and unlike Perlmutter and Postal), I use the following set of primitives: agent-of, subject-of, and direct-object-of.

By taking these grammatical relations as primitives, one can outline means for demonstrating the “sameness” or “differentness” in the syntactic behaviour of subjects, objects, and agents.  That is, based on the expected behaviour of these primitives, one may develop test criteria for determining the degree of ergativity in the syntax of a language.  By observing changes in the state of these relations as other factors in the sentence change, the rules of syntax are revealed.  One may discover the syntactic rules of Suleimaniye Kurdish by applying such syntactic test criteria to Suleimaniye Kurdish, and then observing which of the theoretically-possible combinations of grammatical relationships are allowable.

In the following section, I discuss four syntactic rules which may be used to determine the degree of ergativity in the syntax of a language.  These rules are:  Raising; Equi-NP Deletion; Conjunction Formation; and Relative Clause Formation.  In the sections that follow, I describe each rule, apply it to Suleimaniye Kurdish, and discuss the results.  Since no languages have yet been found which exhibit pure ergative structures and syntax, these tests should be considered to be indicative, rather than conclusive.

3.1  Raising

Raising rules move noun phrases from a subordinate clause into a main clause.  The rule which yields sentence 5b from sentence 5a is an example of such a rule, i.e., one which ‘raises’ the subject of the second clause into that of the first.  In English, for example, some verbs permit subjects and agents to raise,  and others permit objects to raise. Typical main-clause verbs which permit raising in many languages are ‘seem’, ‘be possible’, ‘be easy’, and ‘be able’.

In order to use raising rules to test for ergativity, one must first examine the syntax and lexicon to discover which verbs permit raising.  Among this set of verbs, one must then check to see which of the grammatical relations subject, agent, and object, can be raised.

If subjects and objects in the language under investigation could be raised, but agents could not, this would be evidence of ergative/absolutive syntax.  If subjects and objects cannot be raised, but agents can, this would also be evidence of ergativity.  If, however, subjects and agents can be raised, but objects cannot, or objects and subjects can but agents cannot, this would be evidence of nominative/accusative syntax.  If objects, subjects, and agents may all be raised, then this test will not be useful for determining the presence of ergativity.

It will be noted in the examples from Suleimaniye Kurdish of raising and equi-NP-deletion that the verb of the embedded clause remains finite after the application of the rule.  This would seem to violate Chomsky’s “Tensed-S Condition” [ Chomsky 1973:238 ], which states that rules cannot apply to both matrix and embedded clauses; thus, according to the Tensed-S Condition, as a result of the application of Raising and Equi-NP-deletion, the verb in the embedded clause must become an infinitive form. However, in Suleimaniye Kurdish there are no transformations which yield infinitive verbal forms in embedded clauses.  It should be noted that verbs in embedded clauses, while not infinitives and though they are marked for person and number, are nonetheless in a rather tenseless form, i.e., the subjunctive mood.  Furthermore, infinitives in Suleimaniye Kurdish occur only as nominals.  I would agree, based on these data from Suleimaniye Kurdish, with Barjasteh [ 1983:82-84 ] that Chomsky’s Tensed-S Condition is too strong as stated and that, given the counter-evidence from Persian and Suleimaniye Kurdish, it is unacceptable as a universal principle.  Barjasteh’s first argument is based on the fact that in some cases in Persian, the verb of the embedded clause may be either in the indicative mood or the subjunctive;  his claim is that though this difference results in a change in meaning, the underlying representation is the same.  His second argument is based on the fact that the possessive reflexive rule in Persian may operate across clauses.

Of the verbs in Suleimaniye Kurdish which do permit raising, different verbs have different restrictions concerning which noun phrases may undergo raising.   The  verbs  I  will discuss here are  /wâ dyâr bûn/ ‘seem’, /lawa çûn/ ‘seem’, and /âsân bûn/ ‘be easy’.

The verb /wâ dyâr bûn/ allows object raising, as shown in present tense example 6a.  Note that a clitic copy of the object remains on the verb.  In the past tense, however, object raising is not possible for /wâ dyâr bûn/, example 6b.  Subject raising is possible with /wâ dyâr bûn/ in the present tense, but not in the past tenses, as shown in examples 6c, d, and e.  However, as with nearly all Suleimaniye Kurdish verbs which permit raising, the unraised form of /wâ dyâr bûn/ is preferred over any raised form, as in the example 6f.  The verb /wâ dyâr bûn/ does not permit agents to raise, neither in present nor in past tenses, as in examples 6g and h.

Thus, the verb /wâ dyâr bûn/ permits objects and subjects to raise in the present tense, but not in the past tenses.  This verb does not permit agents to raise at all, in any tense.  In all cases, the unraised variant is preferable to the raised variant.  This would seem to indicate some kind of ergativity, though not what one might expect from the results of the morphological tests; being that morphological ergativity occurs in Suleimaniye Kurdish in the past tenses rather than in the present, one might expect to find syntactic ergativity in the past, rather than in the present, as seems to be occurring here in the examples of /wâ dyâr bûn/.

The verb /lawa çûn/, ‘seem’, permits objects (example 7a and b), subjects (example 7c and d), and agents ( examples 7f) to raise.  However, the last two, though possible, sounded clumsy to my informant, who preferred no raising at all (example 7g) to subject and agent raising.  Thus, though the unraised form is the preferred form, the verb /lawa çûn/ permits objects, subjects and agents to raise, in all tenses, and is therefore not indicative of any preference for either ergative/absolutive or nominative/accusative syntax.

The verb /âsân bûn/, ‘be easy’, permits object raising in both present and past tenses (examples 8a and b), but neither agent nor subject raising (examples 8c, d, e, and f).  This is clearly a nominative/accusative pattern.

Summary of Raising:  To summarize the findings regarding raising: each verb in Suleimaniye Kurdish which was tested has its own restrictions on raising.  There are verbs in Suleimaniye Kurdish which permit raising of all three grammatical relations (subject, agent, and object) in both the present and past tenses (e.g., /lawa çûn/).  There are verbs which may permit only one or two of these categories to raise; for example,  /âsân bûn/ permits only objects to raise.  One of the examined verbs was found (/wâ dyâr bûn/) was found to permit only subjects and objects to raise, and only in the present tenses.  This latter example would seem to indicate that the categories of object and subject were being treated as a single category in this case.  However, since each verb seems to have its own restrictions, it seems more likely that the evidence of /wâ dyâr bûn/ is peculiar to this verb, and not indicative of any broader implications for ergative/absolutive syntax.  The examination of the rules for raising in Suleimaniye Kurdish are therefore inconclusive as a means for determining the degree of ergativity present in the grammar of Suleimaniye Kurdish.

3.2  Equi-NP Deletion

In a sentence with two clauses, where there is a noun phrase in the matrix clause which is co-referential to a noun phrase in the embedded clause, the noun phrase in the embedded clause may delete.  This rule is referred to as “Equi-Noun-Phrase Deletion”.  In example 9, the subject of the embedded clause (“Ali leave”) is identical to the subject of the matrix clause (“Ali wishes”); therefore the subject of the embedded clause is deleted.
 
Equi-NP Deletion rules can be used to determine the extent of ergativity in a language by noting the following two types of noun-phrase behaviour:  (1) which of the grammatical relations subject, agent, or object, control deletion of co-referential noun phrases in embedded clauses, and (2) which of the grammatical relations in embedded clauses may delete.  If subjects and objects in the matrix clause control deletion, but agents do not, or if agents in the matrix clause control deletion while subjects and objects cannot (an unlikely possibility), then this might be evidence of ergative/absolutive syntax.  If, on the other hand, subjects and agents control deletion while objects do not, or if only objects control deletion, while subjects and agents do not, then this might be evidence of nominative/accusative syntax.   Also, if there is a difference in behaviour of noun phrases in the embedded clause, such that subjects and objects in embedded clauses may delete and agents may not, or only agents may delete and subjects and objects may not, then this might also be evidence of ergative/absolutive syntax.  However, if in embedded clauses subjects and agents may delete and objects may not, or objects may delete and subjects and agents may not, thus might be evidence of nominative/accusative syntax.

As was noted in the discussion on Raising, the evidence from Suleimaniye Kurdish regarding Equi-NP Deletion seems to be a counter-example of Chomsky’s Tensed-S Condition.  Equi-NP Deletion results, like Raising, not in an infinitive form, but in a finite subjunctive form which is marked for person and number.

In Suleimaniye Kurdish, as in English, verbs differ with regard to which grammatical relations in matrix clauses may control deletion of co-referential noun phrases in embedded clauses, and which grammatical relations in embedded clauses may delete.  I will discuss here two verbs in Suleimaniye Kurdish which have different restrictions on Equi-NP Deletion, /wâ lê kirdin/, “to induce”, and /wistin/, “to want”.

With the verb /wâ lê kirdin/, “to induce”, in Suleimaniye Kurdish, the direct object of the matrix clause is the only relation which controls deletion of co-referential noun phrases in the embedded clause.  This is true regardless of tense, as the examples in 10 illustrate.

When the clauses  [John induces George / George leave] are collapsed into one sentence, it is the object of the verb of the first clause which controls deletion (example 10a).  This is true in the past tense, as well (example 10b).  The verb “induce” cannot have either subjects or agents controlling deletion.  In Suleimaniye Kurdish, it is not even possible to form ungrammatical approximations of such sentences.

Of the possibilities for which relations in embedded clauses may delete, with /wâ lê kirdin/ only subjects and agents may delete; objects may not delete.  In the following examples, the agent of the embedded clause deletes, in both the present and past tenses (examples 10c,and d).  In example 10e, embedded objects which are co-referential with a noun phrase in the main clause can never be deleted.  Despite the fact that both the agent and the object of the second clause have been deleted, the sentence can only mean that John has induced George to hit John, not that John has induced George (in order) to hit George.

The verb /wistin/, “to want”, in Suleimaniye Kurdish has different restrictions on the application of Equi-NP Deletion.  With this verb, all three relations, object, subject, and agent, may control deletion in the embedded clause.  All three, when in embedded clauses and co-referential to noun phrases in the matrix clause, may delete; all three leave behind clitics on the verb in the embedded clause.

All three relations may control deletion regardless of tense, as the examples in 11. illustrate.  In examples 11 a and b, the direct object of the matrix clause controls deletion.  Examples 11c and d illustrate control of deletion by subjects.  In examples 11e and f, deletion is controlled by the agent of the matrix clause.

Examples 11g-k illustrate the deletion of all three relations in embedded clauses.  In examples 11g and h, objects in embedded clauses delete, in both the past and present tenses.  In examples 11i and j, subjects of embedded clauses delete.  In examples 11k and l, agents of embedded clauses delete.

Summary of Equi-NP Deletion:  The evidence from Equi-NP Deletion seems to indicate nominative/accusative syntax for Suleimaniye Kurdish, rather than ergative/absolutive.  The verb /wâ lê kirdin/ obviously treats subject and agent as a single category, with these two relations able to delete, but not to control deletion, and the relation of object able to control deletion, but not to delete.   The verb /wistin/ treats all three relations alike, with all able to delete and to control deletion, and therefore is not evidence one way or the other.  Other verbs have different restrictions on Equi-NP Deletion, but the result is the same, that subject and agent are treated as a single category, and object is treated as a separate category.

3.3  Conjunction Formation

In a sentence with two clauses which are joined by a conjunction, and where both clauses have co-referential noun phrases, the co-referential noun phrase from one of the clauses may delete.  This is true, however, “only when this [shared] material fills the same syntactic role in both [clauses]” [ Anderson 1976:9 ].  In (hypothetical) example 12a, two clauses with co-referential subjects are joined by a conjunction and the second subject deletes.

When the second clause is transitive, there is the theoretical possibility that either the agent or the object of the second clause, if co-referential to the subject or the agent of the first clause, may delete.  The agent may be delete, as in (hypothetical) example 12b.  It is theoretically possible that the object may delete, as in (hypothetical) example 12c.

If the object of the second clause deletes when it is co-referential to the subject of the first clause (as in example 12c), then, according to Anderson’s statement, the object of the second clause must be interpreted as filling the same syntactic role as the subject of the first clause.  Thus, a syntactically ergative conjunction structure might be indicated if subjects and objects of a second clause which are co-referential to subjects or objects of a first clause delete.   If subjects and agents of the second clause which are co-referential to subjects or agents of the first clause delete, this would indicate a syntactically nominative/accusative conjunction structure.

The evidence from Suleimaniye Kurdish seems to indicate that in the formation of conjunctions, Suleimaniye Kurdish follows a nominative/accusative pattern.  Where two clauses in Suleimaniye Kurdish with co-referential noun phrases are joined by a conjunction, subjects or agents of second clauses which are co-referential to subjects or agents in the first clause may delete; but objects of the second clause which are co-referential to subjects of the first clause may not.

Examples 13a and b illustrate the deletion of embedded-clause agents and subjects which are co-referential to subjects or agents in the first clause, in the present tense.  In the past tense (examples 13c and d), both subjects and agents in second clauses may delete.

The ungrammatical sentences in 14 demonstrate that object deletion is not possible for embedded-clause objects which are co-referential to subjects or agents in the first clause in sentences joined by conjunctions in Sulemaniye Kurdish.   A sentence of this type would only be possible with a pronoun (‘him’) in the second clause, referring to ‘Ali’ (example 14b).  The same is true in the past tense (example 14c).  Attempts  to delete a second-clause object are ungrammatical (example 14d).
 
Summary of Conjunction Formation:  To review the rules in Suleimaniye Kurdish regarding noun phrase deletion in the formation of conjoined clauses:  in a conjunction of two sentences with co-referential noun phrases, subjects and agents may be deleted, but objects may not.  Thus, the rule of conjunction formation in Suleimaniye Kurdish treats the relations of subject and agent as a single category, and treats the relation of object differently.  This is indicative of a nominative/accusative syntactic structure.

3.4  Relative Clause Formation

Relative clauses are formed by embedding one clause into another, when both clauses share a co-referential noun phrase.  This shared noun phrase then deletes, and a pronoun (‘which’, ‘who’, or ‘whom’) or conjunction (‘that’) is inserted in its place.  For example, 15b is derived from 15a.

In a sentence with an embedded relative clause, three possibilities as to which relations may relativize (i.e., delete and be replaced) are subject, object, or agent.  However, there may be restrictions on which relations in the embedded clause may relativize. If only subjects and agents may relativize, and objects may not, or objects may relativize and subjects and agents may not, then this might be evidence of nominative/accusative syntax.  If subjects and objects may relativize and agents may not, or if agents may relativize while subjects and objects may not, then this might be considered evidence of ergative/absolutive syntax.  If all three may relativize (which is true for English), then this test is not useful for determining whether or not a language employs ergative/absolutive or nominative/accusative syntax.

If all three types of noun phrases may relativize, distinguishing features might then be found in different strategies for relativizing.  For example, if there is a difference in the use of ‘who’, ‘which’, and ‘that’ based on the role of the noun phrase in the second clause, this might be evidence of ergativity; or the difference may be a matter of overt marking versus O/-marking.

In Suleimaniye Kurdish, all three relations, subject, agent, and object, may relativize.  Further, there is only one strategy for forming relative clauses, i.e., deletion of the co-referential noun phrase and optional insertion of the conjuction /ka/ “that”.  In examples 16a and b, subjects relativize.  In examples 17a and b, the agent relativizes.  Examples 18a and b illustrate the relativization of objects.
 
The conjunction /ka/ is entirely optional, and may be present or absent with equal correctness.  It is the only relative pronoun or conjunction in Suleimaniye Kurdish.  Thus, regardless of tense, subjects, objects, and agents all relativize in the same manner.  Therefore, relativization is not applicable for the purpose of determining the presence of ergative syntax.

3.5  Summary of Syntactic Test Criteria

I will now summarize the results of the application of the syntactic test criteria to the grammar of Suleimaniye Kurdish.

The application of the Raising tests do not show any conclusive evidence of ergative syntax in Suleimaniye Kurdish.  It appears that rules regarding raising in Suleimaniye Kurdish are different for each verb.  Of the verbs examined in this work, one allowed all three categories to raise, thus being useless for the purposes of this study.  Another verb allowed only objects to raise, thus being an example of nominative/accusative syntax.  One verb permitted only subjects and objects to raise, but only in the present tense.  Since it is most likely that any ergativity present in Suleimaniye Kurdish will be manifest in the past tenses, and not in the present tense, and because this is the only counter-example to what seems to be overwhelming evidence against the presence of ergative syntax in Suleimaniye Kurdish, the evidence provided by this verb is probably only the result of the individual restrictions of this verb, and not an indication of any general trend in Suleimaniye Kurdish syntax.

The evidence regarding the Equi-NP Deletion tests demonstrate a syntactic structure which follows nominative/accusative patterns, since subjects and agents may delete, but objects may not.

The evidence from the application of the Conjunction Formation tests indicate nominative/accusative syntax for Suleimaniye Kurdish, since in the process of conjunction formation, embedded-clause subjects and agents may delete, but embedded-clause objects may not.

The data from the Relative Clause Formation tests give no evidence one way or the other with regard to ergative/absolutive versus nominative/accusative syntax, since all three relations subject, agent, and object are treated alike, i.e., all three may relativize, and in the same manner.

Thus, the results of applying the syntactic test criteria for ergativity to Suleimaniye Kurdish are either ambiguous, or demonstrate a nominative/accusative syntax.

4.  SUMMARY and CONCLUSION

In reviewing the combined results of the application of morphological and syntactic test criteria to Suleimaniye Kurdish grammar, one must conclude that the only ergative rules in Suleimaniye Kurdish are those in the morphology, and only in the past tense; none are strongly present in the syntax, in any tense.  It might be possible to find syntactic structures (like the examples discussed in “Raising”) in Suleimaniye Kurdish which treat subject and object as a single category, and agent differently; but like such structures which Moravscik [ 1978a ] has found in English and other generally nominative/accusative languages, these should be considered as part of the variety of structures which occur in every language, and not used to make general characterizations of the syntax.