Ergative–absolutive alignment

If the language has morphological case, then the verb arguments are marked thus:

• The agent of a transitive verb (A) is marked as ergative case, or as a similar case such as oblique.
• The core argument of an intransitive verb (S) and the object of a transitive verb (O) are both marked with absolutive case.[3]

If there is no case marking, ergativity can be marked through other means, such as in verbal morphology. For instance, Abkhaz and most Mayan languages have no morphological ergative case, but they have a verbal agreement structure that is ergative. In languages with ergative–absolutive agreement systems, the absolutive form is usually the most unmarked form of a word (exceptions include Nias and Tlapanec).[6]

The following examples from Basque demonstrate an ergative–absolutive case marking system:

Here -Ø represents a zero morpheme, as the absolutive case is unmarked in Basque. The forms for the ergative are -k after a vowel, and -ek after a consonant. It is a further rule in Basque grammar that in most cases a noun phrase must be closed by a determiner. The default determiner (commonly called the article, which is suffixed to common nouns and usually translatable by "the" in English) is -a in the singular and -ak in the plural, the plural being marked only on the determiner and never the noun. For common nouns, this default determiner is fused with the ergative case marker. Thus one obtains the following forms for gizon ("man"): gizon-a (man-the.sing.abs), gizon-ak (man-the.pl.abs), gizon-ak (man-the.sing.erg), gizon-ek (man-the.pl.erg). When fused with the article, the absolutive plural is homophonous with the ergative singular. See Basque grammar for details.[7]

In contrast, Japanese is a nominative–accusative language:

In this language, the argument of the intransitive and agent of the transitive sentence are marked with the same nominative case particle ga, while the object of the transitive sentence is marked with the accusative case o.

If one sets: A = agent of a transitive verb; S = argument of an intransitive verb; O = object of a transitive verb, then we can contrast normal nominative–accusative English with a hypothetical ergative English:

Accusative English:

He (A) found me (O).

He (S) traveled.

(S form = A form)

Hypothetical ergative English:

He (A) found me (O).

Him (S) traveled.

(S form = O form)

A number of languages have both ergative and accusative morphology. A typical example is a language that has nominative-accusative marking on verbs and ergative–absolutive case marking on nouns.

Georgian has an ergative alignment, but the agent is only marked with the ergative case in the perfective aspect (also known as the "aorist screeve"). Compare:

K'ac'i vašls č'ams. (კაცი ვაშლს ჭამს) "The man is eating an apple."

K'ac'ma vašli č'ama. (კაცმა ვაშლი ჭამა) "The man ate an apple."

K'ac'- is the root of the word "man". In the first sentence (present continuous tense) the agent is in the nominative case (k'ac'i ). In the second sentence, which shows ergative alignment, the root is marked with the ergative suffix -ma.

However, there are some intransitive verbs in Georgian that behave like transitive verbs, and therefore employ the ergative case in the past tense. Consider:

K'ac'ma daacemina. (კაცმა დააცემინა) "The man sneezed."

Although the verb "sneeze" is clearly intransitive, it is conjugated like a transitive verb. In Georgian there are a few verbs like these, and there has not been a clear-cut explanation as to why these verbs have evolved this way. One explanation is that verbs such as "sneeze" used to have a direct object (the object being "nose" in the case of "sneeze") and over time lost these objects, yet kept their transitive behavior.

In rare cases, such as the Australian Aboriginal language Nhanda, different nominal elements may follow a different case-alignment template. In Nhanda, common nouns have ergative-absolutive alignment—like in most Australian languages—but most pronouns instead follow a nominative-accusative template. In Nhanda, absolutive case has a null suffix while ergative case is marked with some allomorph of the suffixes -nggu or -lu. See the common noun paradigm at play below:[8]

Intransitive Subject (ABS)

Transitive Subject-Object (ERG-ABS)

Compare the above examples with the case marking of pronouns in Nhanda below, wherein all subjects (regardless of verb transitivity) are marked (in this case with a null suffix) the same for case while transitive objects take the accusative suffix -nha.

Intransitive Pronoun Subject (NOM)

Transitive Pronoun Subject-Object (NOM-ACC)

Ergativity may be manifested through syntax, such as saying "Arrived I" for "I arrived", in addition to morphology. Syntactic ergativity is quite rare, and while all languages that exhibit it also feature morphological ergativity, few morphologically ergative languages have ergative syntax. As with morphology, syntactic ergativity can be placed on a continuum, whereby certain syntactic operations may pattern accusatively and others ergatively. The degree of syntactic ergativity is then dependent on the number of syntactic operations that treat the subject like the object. Syntactic ergativity is also referred to as inter-clausal ergativity, as it typically appears in the relation of two clauses.

Syntactic ergativity may appear in:

• Word order (for example, the absolutive argument comes before the verb and the ergative argument comes after it)
• Syntactic pivots
• Relative clauses – determining which arguments are available for relativization
• Subordination
• Switch reference

Example of syntactic ergativity in the "conjunction reduction" construction (coordinated clauses) in Dyirbal in contrast with English conjunction reduction. (The subscript (i) indicates coreference.)

English (SVO word order):

1. Father returned.
2. Father saw mother.
3. Mother saw father.
4. Father_(i) returned and father_(i) saw mother.
5. Father_(i) returned and _____(i) saw mother.
6. Father_(i) returned and mother saw father_(i).
7. * Father_(i) returned and mother saw _____(i). (ill-formed, because S and deleted O cannot be coreferential.)

Dyirbal (OSV word order):

1. Ŋuma banagan^yu. (Father returned.)
2. Yabu ŋumaŋgu buṛan. (lit. Mother father-ŋgu saw, i.e. Father saw mother.)
3. Ŋuma yabuŋgu buṛan. (lit. Father mother-ŋgu saw, i.e. Mother saw father.)
4. Ŋuma_(i) banagan^yu, yabu ŋumaŋgu_(i) buṛan. (lit. Father_(i) returned, mother father-ŋgu_(i) saw, i.e. Father returned, father saw mother.)
5. * Ŋuma_(i) banagan^yu, yabu _____(i) buṛan. (lit. *Father_(i) returned, mother _____(i) saw; ill-formed, because S and deleted A cannot be coreferential.)
6. Ŋuma_(i) banagan^yu, ŋuma_(i) yabuŋgu buṛan. (lit. Father_(i) returned, father_(i) mother-ŋgu saw, i.e. Father returned, mother saw father.)
7. Ŋuma_(i) banagan^yu, _____(i) yabuŋgu buṛan. (lit. Father_(i) returned, _____(i) mother-ŋgu saw, i.e. Father returned, mother saw father.)

Crucially, the fifth sentence has an S/A pivot and thus is ill-formed in Dyirbal (syntactically ergative); on the other hand, the seventh sentence has an S/O pivot and thus is ill-formed in English (syntactically accusative).

The term ergative–absolutive is considered unsatisfactory by some, since there are very few languages without any patterns that exhibit nominative–accusative alignment. Instead they posit that one should only speak of ergative–absolutive systems, which languages employ to different degrees.

Many languages classified as ergative in fact show split ergativity, whereby syntactic and/or morphological ergative patterns are conditioned by the grammatical context, typically person or the tense/aspect of the verb. Basque is unusual in having an almost fully ergative system in case-marking and verbal agreement, though it shows thoroughly nominative–accusative syntactic alignment.[9]

In Hindustani (Hindi and Urdu), the ergative case is marked on agents in the perfective aspect for transitive and ditransitive verbs (also for intransitive verbs when they are volitional),[10] while in other situations agents appear in the nominative case.

In the Northern Kurdish language Kurmanji, the ergative case is marked on agents and verbs of transitive verbs in past tenses, for the events actually occurred in the past. Present, future and "future in the past" tenses show no ergative mark neither for agents nor the verbs. For example:

(1) Ez diçim. (I go)

(2) Ez wî dibînim. (I see him.)

(3) Ew diçe. (He goes)

(4) Ew min dibîne. (He sees me.)

but:

(5) Ez çûm. (I went)

(6) Min ew dît. (I saw him.)

(7) Ew çû. (He went.)

(8) Wî ez dîtim. (He saw me.)

In sentences (1) to (4), there is no ergativity (transitive and intransitive verbs alike). In sentences (6) and (8), the ergative case is marked on agents and verbs.

In Dyirbal, pronouns are morphologically nominative–accusative when the agent is first or second person, but ergative when the agent is a third person.

Many languages with ergative marking display what is known as optional ergativity, where the ergative marking is not always expressed in all situations. McGregor (2010) gives a range of contexts when we often see optional ergativity, and argues that the choice is often not truly optional but is affected by semantics and pragmatics. Unlike split ergativity, which occurs regularly but in limited locations, optional ergativity can occur in a range of environments, but may not be used in a way that appears regular or consistent.

Optional ergativity may be motivated by:

• The animacy of the subject, with more animate subjects more likely to be marked ergative
• The semantics of the verb, with more active or transitive verbs more likely to be marked ergative
• The grammatical structure or [tense-aspect-mood]

Languages from Australia, New Guinea and Tibet have been shown to have optional ergativity.[11]

• Chibchan languages
• Chinookan languages (extinct)
• Coosan languages[12] (extinct)
• Eskimo–Aleut languages
• Guaicuruan languages
• Macro-Jê languages
• Mayan
• Mixe–Zoque
• Panoan languages
• Salish languages
• Tsimshian

• Tedaga, a Nilo-Saharan language of Southern Libya
• Majang language, a Nilo-Saharan language of Ethiopia
• Päri, although recent studies imply a nominative-accusative system.[13]

• Assamese
• Burushaski
• Chukchi[14] (endangered)
• Hawu[15]
• Tibetan
• Sylheti
• Yaghnobi
• Pashto

• Most Australian Aboriginal languages, such as Dyirbal

Certain Australian Aboriginal languages (e.g., Wangkumara) possess an intransitive case and an accusative case along with an ergative case, and lack an absolutive case; such languages are called tripartite languages or ergative–accusative languages.

• Eastern Trans-Fly languages
• various Trans–New Guinea languages

• Basque

• Hurrian (extinct)
• Urartian (extinct)
• Sumerian (extinct)[16]
• South Caucasian: Georgian, Laz
• Northeast Caucasian: Chechen, Lezgian, Tsez, Archi (endangered)
• Northwest Caucasian: Abkhaz, Circassian, Ubykh (extinct)
• Kurdish: Gorani,[17] Zazaki,[18][19][20] Sorani[21][22][23][24][25] and Kurmanji[26][27][28]

Several scholars have hypothesized that Proto-Indo-European was an ergative language, although this hypothesis is controversial.[29]

• In Hindi (Indo-Aryan), ergative alignment occurs only when the verb is in the perfective aspect for transitive verbs (also for intransitive verbs but only when they are volitional).
• In Pashto, ergative alignment occurs only in the past tense.
• In Georgian, ergativity only occurs in the perfective.
• The Philippine languages (e.g., Tagalog) are sometimes considered ergative (Schachter 1976, 1977; Kroeger 1993); however, they have also been considered to have their own unique morphosyntactic alignment. See symmetrical voice.
• In the Neo-Aramaic languages, which are generally classified into 4 groups, only Northeastern (NENA) and Ṭuroyo groups exhibit split ergativity, which is formed in the perfective aspect only, whereas the imperfective aspect is nominative-accusative. Some dialects would only mark unaccusative subjects as ergative. Assyrian Neo-Aramaic, in particular, has an ergative type of construction of the perfective past verbal base, where foregone actions are verbalized by a passive construction with the patient being conferred as the grammatical subject rather than by an active construction, e.g. baxta qtile ("the woman was killed by him"). The ergative type of inflection with an agentive phrase has been extended by analogy to intransitive verbs, e.g. qim-le ("he has risen").[30] Aramaic has historically been a nominative-accusative language.[31]

Sign languages (for example, Nepali Sign Language) should also generally be considered ergative in the patterning of actant incorporation in verbs.[32] In sign languages that have been studied, classifier handshapes are incorporated into verbs, indicating the subject of intransitive verbs when incorporated, and the object of transitive verbs. (If we follow the "semantic phonology" model proposed by William Stokoe (1991)[33] this ergative-absolutive patterning also works at the level of the lexicon: thus in Nepali Sign Language the sign for TEA has the motion for the verb DRINK with a manual alphabet handshape च /ca/ (standing for the first letter of the Nepali word TEA चिया /chiya:/) being incorporated as the object.)