Languages vary in how they mark core arguments. Nominative-accusative languages group the subject of transitives with intransitive subjects, aligning against objects. Ergative-absolutive languages group intransitive subjects and transitive objects, aligning against agents. This fundamental typological parameter has profound consequences for syntax, case marking, and agreement patterns, revealing that subject-hood is defined differently across languages.
To get ergativity, start with something you already know: in English, "the teacher praised the student" and "the student arrived" both use subject position the same way. "The teacher" is the agent — the one performing the action. "The student" in the first sentence is the patient — the one receiving it. "The student" in the second sentence is neither agent nor patient in the transitive sense; it is simply the sole participant in an intransitive event. English groups the transitive agent ("the teacher") with the intransitive subject ("the student arrived") — both appear before the verb with no special marking. The object "the student" in the transitive sentence gets a different treatment (accusative in case-marking languages). This is nominative-accusative alignment: it distinguishes subject-like arguments (nominative) from object-like arguments (accusative), and the defining principle is that agents and intransitive subjects are treated the same.
Now consider flipping the grouping. Instead of grouping transitive agents with intransitive subjects, imagine grouping transitive objects with intransitive subjects — and giving the transitive agent its own special marking. This is ergative-absolutive alignment. The transitive agent gets ergative case; the transitive object and intransitive subject both get absolutive case. If English worked this way, "the teacher" in "the teacher praised the student" would carry a special ergative marker, while "the student" in "the teacher praised the student" and "the student arrived" would look identical (absolutive). The ergative marks the argument that actively causes something to happen to another argument; the absolutive marks the argument most directly affected by the event — whether or not there is a causing agent.
This is not arbitrary linguistic variation. The ergative-absolutive system reflects a coherent semantic distinction: it tracks affectedness rather than agenthood. Ergative languages are neither rare nor exotic — Basque, many Australian Aboriginal languages, Tibetan, Georgian, Mayan languages, and dozens more use ergative-absolutive patterning in some or all of their grammar. What's linguistically important is that ergativity reveals "subject" to be a non-universal category. Your thematic roles work taught you to distinguish agent, patient, and theme as semantic roles. English grammatical subject collapses agents and intransitive experiencers into one syntactic slot. Ergative alignment makes a different cut, one that emphasizes affectedness over agency. Studying typology forces you to recognize that the grammatical categories familiar from European languages are local solutions to cross-linguistic problems, not universal features of human language.
A further complication is split ergativity — many languages are not uniformly ergative across all contexts but split along tense, aspect, or person lines. In Hindi-Urdu, perfective aspect triggers ergative alignment while imperfective triggers nominative-accusative. This is not inconsistency; it reflects a semantic connection between ergativity and completed events where an agent caused a change of state in a patient. In ongoing or habitual events, the agent-patient asymmetry is less sharp, and the language shifts accordingly. Studying ergativity thus integrates morphology (case marking), syntax (agreement patterns), typology (cross-linguistic comparison), and semantics (aspect, agency, affectedness) into a single phenomenon — a model for how linguists understand grammatical categories as solutions to communicative pressures rather than as arbitrary conventions inherited without reason.