Ergative-absolutive alignment groups intransitive subjects with transitive objects (both absolutive) and distinguishes transitive agents (ergative). Languages with ergative alignment include Basque, many Austronesian languages, and many Caucasian languages. Ergativity challenges traditional notions of subject and object, showing that grammatical relations are not universal but vary across languages. Ergative systems correlate with other properties like aspect, voice, and word order.
Analyze ergative language data, identifying ergative and absolutive markers on nouns and verbs. Compare predictions of ergative and accusative grammars for the same phenomena.
From your study of ergativity and linguistic typology, you already know that grammatical relations — subject, object, agent — are not universal givens but language-specific organizational choices. Ergative-absolutive alignment takes that insight and makes it concrete through a specific grouping pattern that seems backwards from an English speaker's perspective. The key to understanding it is to ask: what does a given grammatical system treat as "the same"?
In English's nominative-accusative alignment, the subject of an intransitive verb (*she* sleeps) is marked the same way as the agent of a transitive verb (*she* kicked him). The object of a transitive verb (she kicked *him*) gets different treatment. English groups: intransitive-subject + transitive-agent together, transitive-object separately. In ergative-absolutive alignment, the grouping runs differently: intransitive-subject (*she* sleeps) is marked the same as the transitive-object (she kicked *him* → *him* is "absolutive"). The transitive agent (she kicked him → *she* is "ergative") gets a separate marker. Schematically: ergative alignment groups S(intransitive) with O(transitive) in the absolutive case, and isolates A(transitive) in the ergative case.
Why would a language do this? The grouping isn't arbitrary — it correlates with semantic distinctions. The intransitive subject and the transitive object share a property: both are typically patients or at least non-instigators in the event. The sleeping person isn't doing anything to anyone; neither is the kicked person. The transitive agent is the distinctive role — the one actually causing something to happen to someone else. Ergative systems could be interpreted as making agentivity the marked, exceptional role rather than treating it as the default "subject." Languages like Basque, Quechua, and many Australian languages have this property grammaticalized into their case marking.
Split ergativity shows that the pattern is not all-or-nothing. A language might use ergative-absolutive alignment in the past tense (where completed events make the transitive agent more salient) but nominative-accusative in the present or future. Tzotzil, a Mayan language, splits on aspect. Hindi-Urdu splits on tense. This conditionality is itself theoretically significant: it suggests ergativity isn't a single monolithic property but a cluster of related properties that can be distributed unevenly across a grammar. The split also suggests that what looks like a grammatical category is doing discourse-pragmatic work — tracking not just who did what, but how the event is framed in the speaker's perspective.
The broader implication for linguistic typology is that your intuition about what "subject" means, built on English, is unreliable cross-linguistically. The nominative-accusative pattern is common but not universal. Ergative systems represent a genuinely different theory of how events are organized grammatically — one where agency is the exception to be marked rather than the default around which everything else is defined. Analyzing ergative data requires suspending the assumption that every language must have a "subject" in the English sense, and asking instead: what groupings does *this* language make, and what semantic or pragmatic properties does each grouping track?