ERGATIVE CONSTRUCTION The most generally accepted definition of an ergative construction begins with the notion that languages utilize three primitive syntactic relations, referred to as S, A, and O:
S: subject of an intransitive clause (e.g. “I” in “I have arrived”)
A: subject of a transitive clause (e.g. “you” in “You saw me”)
O: object of a transitive clause (e.g. “me” in “You saw me”).
All languages have intransitive clauses containing a verb and one core noun phrase; this noun phrase will by definition be associated with the syntactic relation S. All languages also have transitive clauses containing a verb and two noun phrases; in this case the noun phrase which “could initiate or control the activity” is associated with the syntactic relation A, while the second noun phrase is associated with the syntactic relation O. The actual semantic role associated with the syntactic relation A varies from verb to verb, but is for example the agent for “affect” verbs like hit, the donor for “giving” verbs like give, the speaker for “speaking” verbs like tell, and the perceiver for “attention” verbs like see. An ergative construction is then one in which S has grammatical properties identical to those of O, and distinct from those of A. It is estimated that approximately one quarter of the world’s languages possess ergative constructions. In Iranian, the grammatical properties which may be relevant are case and verb-agreement. If the grammatical property involved is case, then S and O are typically described as being in an “absolute (Abs.)” or “direct” case, while A is in an “oblique (Obl.)” case. By contrast, the more common accusative construction is one in which S has grammatical properties identical to those of A, and distinct from those of O. If the grammatical property involved is case, then S and A are typically described as being in the “nominative (Nom.)” case, and O is in the “accusative (Acc.)” case. To put it differently, in a full ergative construction involving both verb-agreement and case-marking, called by some Iranists also “passive” construction, and used with the past tenses of transitive verbs, the verb accords not with its “agent” or “logical subject” (A), which is put in the oblique case, but with its object (O). This may be illustrated (Table 1) from some Tātī dialects (q.v), where the verb accords in gender and number with the object, when singular in number, and in number, when plural (gender is not distinguished in the plural).
The construction becomes more understandable when we consider the fact that the verb derives from a passive past participle with adjectival morphology; see Yarshater 1962, p. 245 n. 2; 1969, pp. 235 ff.). E. Benveniste (BSL 48, 1952, p. 19) argued that this type of construction was in fact “possessive” and the agent stands in genitive relationship with the verb. It is estimated that approximately one quarter of the world’s languages possess ergative constructions.
In Iranian, the above notions can be further illustrated from Pashto (ergative) and standard Persian (accusative); see Table 2. Pashto is ergative with respect to case-marking, since S and O (zə) are both in the absolute case, while A (tā) is in the oblique case. Pashto is also ergative with respect to verb-agreement, since the verb agrees with S and with O (the inflection being first-person singular -əm in both examples), but does not show any agreement with A. By contrast, Persian is accusative with respect to case-marking, since S and A (man and to respectively) are both in the nominative case, while O has the accusative marking -rā. Persian is also accusative with respect to verb-agreement, since the verb agrees with S and with A (the inflection being first-person singular -am for S and second-person singular -ī for A), but does not agree with O. The ergativity of Iranian languages is restricted to case-marking and verb-agreement, and does not (as far as is known) extend to any other syntactic properties of S, A and O.
It should be noted that Pashto, and ergative Iranian languages generally, exhibit what is known as split ergativity, since the ergative construction is found only in clauses using tenses based on the past stem of the verb. Clauses using tenses based on the present stem are straightforwardly accusative with respect both to case-marking and verb-agreement (the case terms “nominative” and “accusative” case seem inappropriate, however, since the case-forms used are simply the absolute and oblique with reversed roles); see Table 3.
The historical origin of split ergativity in Iranian is well-attested. Whilst the accusative constructions of the present tenses derive straightforwardly from the accusative constructions of Old Iranian, the ergative constructions of the past tenses derive from constructions based on the Old Iranian perfect participle in-ta. As shown in Table 4, this principle had a passive orientation in the case of transitive verbs, with the agent expressed either by a genitive noun phrase (4.i), or a genitive clitic pronoun (4.ii). The participle could also be accompanied by a form of the copula verb ‘ba’ agreeing with the passive subject (4.iii).
The reinterpretation of these constructions as active rather than passive, and the collapse of the genitive and dative into a single oblique case, gave rise to ergative constructions in which the original passive agent was reinterpreted as an A in the oblique case, while the original passive subject was reinterpreted as an O in the absolute case. The verb agreement was then also oriented towards O.
Relatively few of the middle or modern Iranian languages, however, display pure past-tense ergative constructions as defined above and as exhibited in Pashto. Yaḡnōbī and Kormānjī dialects of Kurdish are ergative with respect to both case-marking and verb-agreement. Ṭālešī has ergative case-marking, but has lost ergative verb-agreement, while Middle Persian and Parthian have largely lost ergative case-marking, but preserve ergative verb-agreement. Other languages have typically lost both ergative case-marking and ergative verb-agreement, and display various stages in the decay of the originally ergative past-tense construction into an accusative one. The potential influence of the accusative construction in the present tenses is evident. Standard developments are:
(a) The conversion of the absolute case-marking of O into oblique. If this happens at the same time as oblique case-marking is preserved for A, we get the same marking for A and O (e.g. in Rōšānī), and a distinct marking for S. This has been called the “double-oblique” case-marking system, and is an extremely rare phenomenon. Other possible developments include the further grammaticalization of prepositions or postpositions as object-markers for O (e.g. Persian -rā < Middle Persian rāy < Old Persian rādiy “for the sake of”). If this happens at the same time as oblique case-marking is preserved for A, we get distinct marking for A, O and S (e.g. in Yazḡolāmī). This has been called the “tripartite” case-marking system, and is likewise very rare.
(b) The leveling of the case-forms for A and S into a single form. In this event, the forms chosen may correspond either to the absolute (e.g. Ossete əe “I”) or the oblique (e.g. Persian man “I”). Co-existence of absolute and oblique subjects is also possible, in which case it has been claimed that the choice of case may involve semantic and discourse-related functions (e.g., in the upper dialect of Waḵī; see Bashir).
(c) The development of clitic pronouns such as -šam in (ii) above into agreement markers for A (e.g., Lārī). This typically does not occur simultaneously with preservation of the original agreement with O.
(d) The extension of the function of S agreement markers, typically derived from the copula as in (iii) above, to agreement with A. Such a development seems to have taken place in Persian, where the forms of verb-agreement with S and A are identical. In complicated cases (e.g. many Pamir languages), we see the contamination of agreement markers derived from clitic pronouns with those derived from the copula (with the further possible involvement of endings belonging to the present tense).
As an illustration of the complexity which can arise in the decay of the original ergative construction, reference may be made to a transitive sentence from Bartangī in which case-marking is tripartite, with the object-marker derived from the preposition az|as “from,” and in which clitic pronouns/copulas have been converted into agreement markers of A (Table 5). The agreement marker -um in this example is cliticized not to the verb, but to the fronted and object-marked O.
E. Bashir, “Beyond Split-ergativity: Subject Marking in Wakhi,” in Chicago Linguistic Society, Papers from the General Session at the Twenty-Second Regional Meeting, Chicago, 1986, pp. 14-35.
E. Benveniste, BSL 48, 1952, p. 19.
R. M. W. Dixon, Ergativity, Cambridge, 1994.
Osnovy Iranskogo yazykoznaniya (Fundamentals of Iranian linguistics), Moscow, 1979- .
V. Miltner, “Ergative Constructions in Indo-Aryan,” Archiv Orientalni 59, 1991, pp. 225-33.
J. R. Payne, “The Decay of Ergativity in Pamir Languages,” Lingua 51/2-3, 1980, pp. 147-86.
L.A. Pireyko, Osnovnye voprosy èrgativnosti na materiale indo-iranskikh yazykov (Fundamental issues of ergativity in Indo-Iranian languages), Moscow, 1968.
R. Schmitt, Compendium Lingusrum Iranicarum, Wiesbaden, 1989.
G. Windfuhr, “Typological Notes on Pronominal Cases in Iranian Tati,” Bulletin of the Asia Institute 4, 1990, pp. 313-24.
E. Yarshater, “The Tātī Dialect of Rāwand,” in idem and W. B. Henning, eds., A Locust’s Leg: Papers in Honour of S. H. Taqizadeh, London, 1962, pp. 240-45.
Idem., A Grammar of Southern Tati Dialects, The Hague and Paris, 1969.
(John R. Payne)
Originally Published: December 15, 1998
Last Updated: January 19, 2012
This article is available in print.
Vol. VIII, Fasc. 5, pp. 555-558