Kurdishsplits by tense
Sentences with a doer and a done-to work like English in the present tense, and ergatively in the past.
“She slept.” “She saw him.” Your gut says the sleeper and the seer are the same kind of thing. Plenty of the world’s languages disagree. A few taps, some real Basque, and one tiny ending will show you why.
Here are three people from two tiny sentences. Don’t overthink it: go with your gut.
Tap the two that feel like “the same kind of thing”: the sleeper (“She slept”), the one who sees (“She saw him”), or the one who is seen (“She saw him”).
No wrong answer here. Hold on to your pick. Many English speakers pick the sleeper and the seer, because English uses “she” for both. By the end you’ll meet languages whose grammar draws the line the other way, grouping the lone participant with the one who is seen.
Let’s put those three people under a lens. Scroll on, and watch the canvas as you go.
Start with who’s in each sentence. She slept has one participant, the sleeper, and nobody else. She saw him has two: one who does the seeing, and one who gets seen.
Linguists give these three roles one-letter names. They’ve just appeared on the canvas.
S is the single participant of a one-participant verb like slept.
A is the agent: the doer of a two-participant verb like saw.
P is the patient: the one the action lands on. (Some books call it O, for object.)
Now look at the words themselves. S and A are both she. Only P gets a different form: him, or her. So English draws its line around S and A, and leaves P on its own.
This grouping is called nominative-accusative, or accusative for short.
Because “subject” already bundles S and A together, and that bundle is exactly what this page puts on trial. Three separate letters let us ask which two a language groups.
Also, English shows the grouping in its word forms only on pronouns. A noun like the man looks the same whether he sleeps, sees, or is seen.
“He fell.” and “He pushed her.” Which pronoun is S, which is A, which is P, and which two share a form?
“Fell” has one participant, so “He” is S. In “pushed”, the pusher “He” is A and the pushed “her” is P. English gives S and A the same form (“he”) and P a different one (“her”). That is accusative alignment. B swaps the doer and the one acted on: the role comes from who does what, not from word order or position alone.
English puts S with A. But is that the only sensible line to draw? Imagine an English that groups differently: it puts the sleeper together with the one who is seen, and leaves the seer on its own.
Before the canvas changes, make a prediction.
Imagine an English that groups the sleeper with the one who is seen, not with the seer. How would “She slept” come out?
If the sleeper is grouped with the one who is seen, the sleeper takes the seen one’s form: “her”. A is real English, which groups the sleeper with the seer. C adds a second participant, so it is no longer the lone-sleeper sentence. Real ergative languages usually don’t swap pronouns like this. They mark the doer with an extra ending, as the next section shows.
Answer the question above, and the canvas will flip.
The outline jumped. In this English the sleeper takes the object form, the one P gets: Her slept. Now S and P share a form, and A, the seer, is the odd one out.
Use the switch on the canvas to flip between the two Englishes. Same three roles; only the grouping moved.
Nobody speaks this English. Real languages that group S with P don’t usually reshuffle pronouns like this. They leave S and P bare, and give the doer an extra ending. Here’s one.
Two Basque sentences: Martin etorri da, ‘Martin has arrived’, and Martinek Diego ikusi du, ‘Martin has seen Diego’.
Same Martin in both. When he simply arrives, his name has no ending. When he does something to someone, he becomes Martin-ek. And Diego, the one seen, stays bare, just like Martin who arrived.
That’s the whole trick. In Basque, the doer of a two-participant verb (A) gets the ending -k, spelled -ek after a consonant, as in Martinek. This ending is called the ergative.
The one acted on (P) stays bare, and so does the lone participant (S) of verbs like ‘arrive’ and ‘come’. The bare form is called the absolutive. S goes with P, and A stands apart: Basque is an ergative-absolutive language.
The canvas switch now flips between all three views. Compare where the outline sits in English and in Basque.
One pair of sentences could be a fluke. Here is one noun, umea ‘the child’, in all three roles.
Good catch, and it’s a real trap. The -a on these nouns works like “the”. Add the doer’s -k and you get gizonak, ‘the man’ as a doer. But gizonak is also how Basque says ‘the men’ in the plain, unmarked form. Only the rest of the sentence tells you which one you’re looking at.
That’s why the puzzles below promise that every noun is singular.
Your turn. Each puzzle gives you an English sentence and the Basque words to say it with. Tap a noun to put the ending on or take it off, then press Check. The puzzles unlock one at a time.
Someone says “Martinek Diego ikusi du” is really a passive, like “Diego was seen by Martin”, since Martin gets the extra marking. Are they right?
It’s the normal way to say “Martin has seen Diego”. In an English passive the doer is optional and demoted (“Diego was seen”). In Basque, -k is simply what every transitive doer wears. Linguists do note a resemblance between ergative patterns and passives, which is why A is tempting, but the Basque sentence is active. C mixes in split ergativity, which is coming up next and isn’t how Basque works.
Some English grammar books call verbs like break “ergative verbs”, because they work both ways: it broke the window and the window broke. That’s a different use of the same word, and linguists haven’t settled on one name for those verbs.
Having verbs like break doesn’t make English ergative. On this page, “ergative” means an ending that the doer of a two-participant verb wears, as in Basque.
So far it might sound as if every language picks one grouping and sticks to it. Many don’t. Hindi-Urdu has its own doer ending, -nē, and it’s a good place to watch a language change its mind.
Hindi-Urdu has an ergative ending, -nē. “The boy sells a book.” vs “The boy has sold a book.” Where does -nē show up on the boy?
Hindi-Urdu marks the agent with -nē only in the perfective (“has sold”). In “sells” the boy is unmarked, as in an accusative language. One language, two alignments, switched by tense/aspect: split ergativity. If you chose A, you assumed a language is either ergative or not. Many are ergative only part of the time.
Now try the switch below.
Same boy, same book. Only the tense changed (strictly, the aspect), from ‘sells’ to ‘has sold’, and the doer’s ending switched on with it. A language that is ergative in some tenses and accusative-looking in others has split ergativity.
That clears up a popular claim: that Kurdish and Georgian are ergative languages, just like Basque. Each is only partly ergative, and each splits in its own way. Dyirbal, from Australia, splits along a different line altogether: by the kind of word.
Sentences with a doer and a done-to work like English in the present tense, and ergatively in the past.
Its ergative ending (called the “narrative” case) shows up mainly in the aorist. Even there, some one-participant verbs take it too, so it ends up sorting verbs by how agent-like their participant is.
Ordinary nouns follow the ergative pattern, but the words for “I” and “you” follow the accusative one, as English pronouns do.
Sources: Wikipedia, “Split ergativity” (Hindi-Urdu, Dyirbal), “Kurdish grammar”, “Georgian grammar”.
Is ergativity a rare oddity, then? The World Atlas of Language Structures (WALS) checked how 190 languages mark these roles on ordinary nouns.
Just over half the languages, 98 of 190, don’t mark these roles on their nouns at all. Among the 92 that do, the accusative grouping is more common, 52 to 32. But ergative is no curiosity: it’s about a third of them.
Why isn’t English’s grouping (sleeper with seer) more logical than an ergative language’s (sleeper with seen)?
Give yourself a moment to answer in your head first.
Each grouping picks out something real. English groups by “who the sentence is about, who starts things”: the sleeper and the seer are both the main participant. An ergative language groups by “who the event happens to”: the lone participant and the one acted on are both simply undergoing or being in the event, and the marking goes mainly to the agent of a two-participant verb. In Basque, that’s why the one who arrives and the one who is seen both stay bare, while the one who sees gets -k.
Both use the same three roles. They draw the line in different places, and both are common in the world’s languages: 52 accusative and 32 ergative in the WALS sample, while just over half mark neither on nouns. Your instinct from the start of the page is your language’s architecture, not universal logic.
Want to keep going? Look up the two small bars in the chart, tripartite and active systems, and syntactic ergativity, where rules of the grammar beyond the endings follow the S+P grouping.