The parameter
Three arguments have to be accounted for in any language. S is the single argument of an intransitive verb, the one that walks. A is the more agent-like argument of a transitive verb, the one that hits. P is the more patient-like argument, the one that gets hit. Languages rarely mark all three differently, so each groups two and singles out the third.
Nominative-accusative alignment groups S with A: the walker and the hitter are treated alike, and the one who gets hit is singled out. That is what English, German, Turkish, Russian and almost every language a European translator works with do, which is why the arrangement feels like a fact rather than a choice. Ergative-absolutive alignment groups S with P instead, so the walker and the one who gets hit pattern together while the hitter takes a case of its own, the ergative.
Alignment is a property of constructions rather than of whole languages, which is where the label set oversimplifies: a language may align its case marking on one pattern and its verb agreement on another. Dixon separates morphological from syntactic ergativity and shows that the syntactic kind, where coordination and relativization also follow the S and P grouping, is rare.
How languages vary
- Nominative-accusative: S and A alike, P distinct. English, German, Russian, Turkish, Japanese, Arabic, and most of Europe.
- Ergative-absolutive: S and P alike, A distinct. Basque, West Greenlandic, many Australian and Caucasian languages, and the Mayan family.
- Split ergative: ergative under some conditions and accusative under others. The split follows aspect in Hindi-Urdu and Georgian, and the animacy scale in Dyirbal, where pronouns pattern accusatively and full nouns ergatively.
- Active-stative: S is treated like A or like P according to the semantics of the event. Guaraní and Lakota are standard examples.
The two WALS chapters on alignment measure different things and should not be added together. Comrie's chapter 98, on case marking of full noun phrases, finds in a sample of 190 languages that 98 mark no core argument by case at all, 46 are standard nominative-accusative and 32 ergative-absolutive. Siewierska's chapter 100, on verbal person marking, counts 212 accusative against 19 ergative out of 380. Ergativity is far commoner in case marking than in agreement.
Consequences for translation
Alignment decides which argument is grammatically prominent. Coming from an ergative language into an accusative one, the argument that carried a special case is the one the target treats as ordinary, and the argument that was unmarked becomes the object. A rendering that keeps the surface subject of the source often puts the emphasis where the source did not, and the fault is hard to see because both versions read naturally. This is where voice change stops being a stylistic option: promoting or demoting an argument may be the only way to hold the prominence in place. The parameter enters European work through the languages met at its edges: Basque, Georgian, Kurdish, Hindi-Urdu, Pashto and Tibetan.
Ergative marking is not emphasis. Where an ergative case marks the agent of every transitive perfective clause, it carries no more rhetorical weight than an English subject. Rendering it with a cleft or a fronted agent adds a figure the source did not have; treating it as backgrounded because it looks oblique is the same error in reverse.
Splits look like inconsistency. In a split-ergative language the same participant may be marked one way in a narrative past and another in a present-tense summary of the same events. That is grammar responding to aspect, not a change of perspective, and a target text that mirrors it by switching between active and passive invents a distinction. For a reviser, the question is whether a voice change answers a real difference or merely tracks a split.
Examples
EU>EN
- Source: Martin etorri da. / Martinek Diego ikusi du.
- Target: Martin has arrived. / Martin has seen Diego.
- Comment: Basque leaves the sole argument of the intransitive clause unmarked and adds the ergative suffix to the agent of the transitive one. In English both are unmarked subjects, so a distinction the Basque grammar makes overtly disappears without trace.
KA>EN
- Source: K'aci vašls č'ams. / K'acma vašli č'ama.
- Target: The man is eating an apple. / The man ate an apple.
- Comment: Georgian marks the agent with an ergative suffix in the aorist and leaves it unmarked in the present; English marks the difference on the verb alone. A target text that switches construction as well as tense reports a change of perspective that did not occur. Hindi-Urdu behaves the same way, with the postposition ne on the agent in the perfective and no marking elsewhere.
References
- Dixon, R. M. W. (1994). Ergativity. Cambridge: Cambridge University Press.
- Dixon, R. M. W. (1979). "Ergativity." Language 55, 59–138.
- Comrie, B. (2013). "Alignment of Case Marking of Full Noun Phrases." In Dryer, M. S. & Haspelmath, M. (eds.), The World Atlas of Language Structures Online, Ch. 98. Leipzig: Max Planck Institute for Evolutionary Anthropology.
- Siewierska, A. (2013). "Alignment of Verbal Person Marking." WALS Online, Ch. 100.
- Comrie, B. (1989). Language Universals and Linguistic Typology, 2nd edition. Chicago: University of Chicago Press, Ch. 5.