Oil In My Lamp wrote: ↑26 Mar 2020 22:24
Here is another example. Perhaps it makes more sense:
The tree was felled.
The tree fell.
Would they both be written like below:
fell tree-DEF
fell tree-DEF
I have a hard time understanding ergative grammar.
The answer is... they
could be.
And here's an important question to ask whenever a conlanger finds that two things that are distinguished in English aren't distinguished in their conlang:
why is that a problem?
"The tree fell" and "the tree was felled" both refer to the same physical event - the tree falling over. The only semantic difference is that the latter hints that someone is to blame, without saying who. How often is that hint needed? Remember, you still have the transitive distinct: "Bob felled the tree" (/"the tree was felled by Bob"). And if you DO sometimes need to say that someone's to blame for the tree falling over but you can't or don't want to say who, you can always just say "the tree was felled by someone".
That is, you could distinguish:
fell tree-DEF -
the tree fell/was felled
fell tree-DEF Bob -
the tree was felled by Bob
fell tree-DEF someone -
the tree was felled by someone
What possibility is still semantically excluded?
--------------
---------------------
However, the other answer is: they
don't have to be.
'To fall' and 'to fell' both describe the same event, but English has two different words for them. Etymologically, "to fell" is a causative form, a verb that was originally explicitly marked to increase valency. These days, this causative marking is no longer productive, but is still visible in the form of a small number of verb pairs - fall/fell, sit/set, lie/lay, rise/raise [footnote: the 'true' pair there is rise/rear; 'raise' is borrowed from the Norse form of the causative; but it doesn't matter for our purposes here].
Let's step back a step.
Every verb has, as it were, a number of "slots" around it that nouns can fill. For most verbs, there's either one slot or two slots. These slots aren't interchangeable; we can, as it were, give them different labels. And the verb then determines what semantic role a noun must play to be eligible to take a slot with a given label.
If a verb has one slot, we can call it S.
If a verb has two slots, we can call them A and O.
If we have a verb with two nouns, the 'marking' (on the noun, on the verb, or in word order) shows which is A and which is O.
But if a verb has only one slot, how do we know whether it's S, A or O?
The easiest way is to say that S, A and O all get different markings, so there can never be ambiguity. This is a classic tripartite alignment.
You can also get de facto tripartite alignments in some 'ergative' languages, where one system of marking (eg nominal case) is ergative-absolutive, but another system of marking (eg verbal agreement) is nominative-accusative. We still call these languages 'ergative', because "ergativity" is really often used as a blanket term for "not perfectly nominative-absolutive".
In most languages, S and A are marked the same way. This is nominative-accusative alignment.
In some languages, S and O are marked the same way. This is ergative-absolutive alignment.
So, to get back to your problem...
...if in a (pure) ergative-absolutive language you find a verb with only one noun attached, how do you know whether the noun is in the S slot or the O slot?
Well, that's the wrong way to see it. Because in a pure ergative-absolutive language, the semantics of the S slot are the same as the semantics of the O slot. In "the tree fell" and "the tree was felled", the tree is still doing exactly the same thing! In "the bear danced" and "the bear was danced", the bear is still doing exactly the same thing!
So really the question is: how do you know whether the verb has a non-overt second argument (in the A slot)?
And there are different possibilities.
The first is: you can't, but it doesn't matter. Really, who cares? If the other argument were important, the speaker wouldn't have chosen to make it non-overt!
The second is: there can't be a non-overt noun in the A slot, because this is illegal. All agents must be explicitly present. This is similar to how many transitive verbs in English very strongly demand an overt object - I can't just say "I enjoyed", I need to include an object. I can't easily say "I felled", either.
The third is: you can tell because it's marked on the verb. If the verb has "THIS VERB HAS TWO SLOTS!" attached to it as a marker, then when you see it with only one overt argument, you know that there must be a non-overt argument in the second slot.
Now, this marking can be done in on three 'levels':
- you could do it purely lexically. Every verb is then uniquely specified from birth, as it were, as having either one slot or two slots, and every time you see that verb you know how many slots it should have. In English, when you see "enjoy", you know there's meant to be two slots. When you see "fell", you know there's meant to be two slots. When you see "fall", you know there's meant to be one slot. You have to just learn which verbs are which.
- you could do it through derivation. If English intransitive/causative pairing was still around, it would be a derivational process: you'd take a verb in one category, and you'd attach a marker, and it would move it into the other category. But it wouldn't apply to all verbs, only ones where it made sense to have a pairing, and the semantics might be a little unpredictable sometimes, and there might be multiple different markers used with different verbs. English be- is another example of this process - again, it added an additional argument, but whereas causative ablaut adds an argument with agent/causer semantics, be- adds an argument with patient/recipient/location semantics. So fall > befall (fall on). It should also be said that these markers can go the other way, and reduce the number of slots a verb has.
- you could do it through inflexion. You could make it so that the marker that changes the number (and role) of slots was predictable, and could be added to any verb willy-nilly, without changing the core semantics at all. The English passive is a form of inflexional valency-alteration.
In order to translate English ergatively, I have been converting most active English sentences to passive form before translating them into my ergative conlang. Are there any pitfalls I should be aware of with this practice?
As Ser says, you can't form a passive of an intransitive - an intransitive only has one slot, and a passive creates a verb with only one slot, so it would be superfluous when attached to an intransitive. In English, you can sometimes form what LOOKS like the passive of an intransitive, but what it really is is the passive of a zero-derived bivalent verb. But because this is still a derivational process in English, the semantics end up unpredictable and weird - sometime the result is the passive of a causative (the bear was danced) and sometimes it's the passive of a transitive (the bear was eaten).
I'd also just say that not all bivalent verbs in English have the same semantic roles, so they needn't all be 'translated' into ergative languages the same way. In English "I like the pie", "I" is not really an agent, and "pie" is not really a patient, so you may well find an ergative language in which "I" is in the absolutive and "pie" is in the ergative.