Unaccusative v unergative v ergative verbs - Can you help?

[matsu]

With my conlang, I'm thinking about shifting over my alignment from accusative-nominative to ergative-absolutive, based on the way I'm handling my verbs. But I need a better grasp of the concept of unaccusative and unergative verbs, and the difference between unergative and ergative. I don't find their Wikipedia articles particularly helpful...

It's my understanding that ergative verbs have the same bit between transitive direct object, and intransitive subject. So, for example, "the food is cooking" v. "I'm cooking the food". Here, "food" is serving both roles as the experiencer. Does that make "cooking" an ergative verb? Or does the "same bit" have to be an actor, and not an experiencer? If this is the case, then is "cooking" then an unaccusative verb when intransitive (and the semantic roles therefore define the difference between these two types of verb)?

I understand the difference between unaccusative and unergative better. From what I've learned, an unaccusative verb is either not responsible, or it does not/cannot initiate the action of the verb. So, to use "cooking" again, "The food is cooking" is thus an unaccusative verb because the food does not initiate its own action; a verb like "I die" is the same, because one is typically not responsible for the end result of the process of dying. In contrast, an unergative verb has an agent with an implied object. So, "I'm cooking" is unergative because there is an implied object that is not stated. But then, what's the difference between unergative and ergative verb?

Are my analyses off? Right? How do ergative-absolutive languages use these types of verbs (unaccusative, unergative, ergative)? What about nominative-accusative languages? If you have examples I'd love that...that way I have something to reference notes with. Thanks!

[finlay]

In English we tend to call this "middle voice" - passive-ish meaning with active syntax. You could certainly make your own conlang be this way more generally than it is in English, and call it ergativity.

[Vardelm]

Here's my take:

Accusative Verb: a verb that can be intransitive or transitive and behaves in a standard accusative manner. The intransitive subject has agent-like semantics, and therefore matches the nominative argument of the transitive.

dog.NOM eat.PERF = The dog ate.

dog.NOM eat.PERF bone.ACC = The dog ate the bone.



Ergative Verb: a verb that can be intransitive or transitive and behaves in a standard ergative manner. The intransitive subject has patient-like semantics, and therefore matches the absolutive argument of the transitive.

pie.ABS bake.PERF = The pie baked.

pie.ABS bake.PERF chef.ERG = The pie was baked by the chef.

Note: I used English passive here to put the arguments in the same order as the example. Also, I wanted to show syntactic ergativity as well as morphological. More on that later.



Unergative Verb: a verb that can only be intransitive. The subject has agent-like semantics, and therefore matches the intransitive subject of accusative verbs.

dog.NOM bark.PERF = The dog barked.



Unaccusative Verb: a verb that can only be intransitive. The subject has patient-like semantics, and therefore matches the intransitive subject of ergative verbs.

dog.ABS die.PERF = The dog died.



The most confusing thing about this really is the terminology. Basically, unaccusative verbs are the intransitive-only sibling of ergative verbs, while unergative verbs are the intransitive-only version of accusative verbs.

Every now and then I post a link to this paper, which explains morphological, syntactic, and semantic ergativity. I find that it is VERY helpful in understanding ergativity.

Typology of Ergativity by William McGregor



If you have a language which is absolutely, 100%, purely ergative, then I believe all of the intransitive-only verbs would be unaccusative. All of the transitive verbs would be ergative. The same can be said for a pure accusative language. All of its intransitives would be unergative, and all transitives would be accusative.

As I understand it, no language is going to be absolutely, 100% purely ergative or accusative. Therefore, a naturalistic ergative language will have a majority of its verbs be unaccusative or ergative, but at least some will be unergative and/or perhaps accusative. A naturalistic accusative language will have a majority of unergative & accusative verbs, with some that are unaccusative or ergative.

English is a good example here: it is an accusative language, but it does have examples of all 4 types of verbs. Only a minority of its verbs are unaccusative or ergative.


Also, to my knowledge, morphologically ergative languages are usually (always?) semantically ergative as well. However, of those languages, most are syntactically accusative, making fully ergative languages quite rare. I think the notion of accusative, ergative, unaccusative, and unergative verbs lies mostly within the domain of semantic ergativity. So, if you have a morphologically ergative language, it will almost certainly be semantically ergative and therefore have a majority of unaccusative & ergative verbs.

For comparison, look at a language which is morphologically & semantically ergative, but syntactically accusative. Basque would be an example. A majority of its verbs would probably structure like this:

Unaccusative:

dog.ABS die.PERF = The dog died.


Unergative:

pie.ABS bake.PERF = The pie baked.

chef.ERG bake.PERF pie.ABS = The chef baked the pie.

[Terra]

So, "I'm cooking" is unergative

To be fair, this is only because of pragmatics. If you imagine a bunch of cannibals throwing you into a giant kettle while you're still alive, you could say "I'm cooking!" where "I" is the patient and not the agent.

Unergative Verb: a verb that can only be intransitive. The subject has agent-like semantics, and therefore matches the intransitive subject of accusative verbs.

dog.NOM bark.PERF = The dog barked.

What about "The boss barked orders to his underlings."?

* * *

Anyways, my take:

First, some terms:
- Subject : In a nom/acc language, the nominative. In an abs/erg language, the absolutive.
- Object : In a nom/acc lanugage, the accusative. In an abs/erg language, the ergative.

Nominative/Accusative languages have two types of verbs:
1) When the verb takes an object, the subject is the agent and the object is the patient. When the verb lacks an object, the object/patient is implied; The subject is still the agent.
-- The dog.NOM ate the apple.ACC.
-- The dog.NOM ate (something/somebody).
2) When the verb takes an object, the subject is the agent and the object is the patient. However, when the verb lacks an object, the subject is the *patient*.
-- The man.NOM broke the window.ACC.
-- The window.NOM broke (because of something/somebody).
3) The verb never takes an object, and the subject is the *patient*. (This is really just a special case of type 2. The point is that there's no object, and the subject is the patient.)
-- The man.NOM died (because of something/somebody).

Ergative/Absolutive languages have two types of verbs:
1) When the verb takes an object, the subject is the patient and the object is the agent. However, when the verb lacks an object, the subject is the *agent*.
-- The apple.ABS was eaten by the dog.ERG.
-- By the dog.ABS was eaten (something/somebody).
2) When the verb takes an object, the subject is the patient and the object is the agent. When the verb lacks an object, the object/agent is implied: The subject is still the patient.
-- The window.ABS was broken by the man.ERG.
-- The window.ABS was broken (by something/someone).
3) The verb never takes an object, and the subject is the patient. (Again, this is really just a special case of type 2.)
-- The man.ABS was killed (by something/somebody).

Nom/Acc Patterns:
1)
-- agent, patient
-- agent
2)
-- agent, patient
-- patient

Abs/Erg Patterns:
1) Note how the pattern is the same as nom/acc type 2 with flipped semantic roles.
-- patient, agent
-- agent
2) Note how the pattern is the same as nom/acc type 1 with flipped semantic roles.
-- patient, agent
-- patient

I don't know how the terminology (unergative, unaccusative, etc) is applied to each of these verbs. The important part is to understand the mechanism.

Edit: Corrected some capitalization.

[Vardelm]

What about "The boss barked orders to his underlings."?

Touche. Feel free to offer up another suggestion.

First, some terms:
- Subject : In a nom/acc language, the nominative. In an abs/erg language, the absolutive.
- Object : In a nom/acc lanugage, the accusative. In an abs/erg language, the ergative.

Saying that the noun w/ ergative case is an object in an abs/erg language is stretching the definition of "object" pretty far. If you mean "any verb argument that is not the subject", then yes. Even then, this would only hold true for syntactically ergative languages. A syntactically accusative language would probably consider the ergative noun to be the subject of a transitive sentence.

Nominative/Accusative languages have two types of verbs:

From the list below, you leave out unergative verbs. Even if my example of "barking" above doesn't hold, are you arguing that there is no nom/acc language anywhere that has an intransitive verb in which the subject has agent-like semantics?

1) When the verb takes an object, the subject is the agent and the object is the patient. When the verb lacks an object, the object/patient is implied; The subject is still the agent.
-- The dog.NOM ate the apple.ACC.
-- The dog.NOM ate (something/somebody).

This would be an accusative verb.

2) When the verb takes an object, the subject is the agent and the object is the patient. However, when the verb lacks an object, the subject is the *patient*.
-- The man.NOM broke the window.ACC.
-- The window.NOM broke (because of something/somebody).

Ergative verb.

3) The verb never takes an object, and the subject is the *patient*. (This is really just a special case of type 2. The point is that there's no object, and the subject is the patient.)
-- The man.NOM died (because of something/somebody).

Unaccusative verb.

1) When the verb takes an object, the subject is the patient and the object is the agent. However, when the verb lacks an object, the subject is the *agent*.
-- The apple.ABS was eaten by the dog.ERG.
-- By the Dog.ABS was eaten (something/somebody).

Accusative verb.

This does bring up an interesting point: would an antipassive voice ever be used with such a verb, since the intransitive use in this case would match the typical use of an antipassive?

2) When the verb takes an object, the subject is the patient and the object is the agent. When the verb lacks an object, the object/agent is implied: The subject is still the patient.
-- The window.ABS was broken by the man.ERG.
-- The window.ABS was broken (by something/someone).

Ergative verb.

3) The verb never takes an object, and the subject is the patient. (Again, this is really just a special case of type 2.)
-- The man.ABS was killed (by something/somebody).

Unaccusative verb.

I don't know how the terminology (unergative, unaccusative, etc) is applied to each of these verbs. The important part is to understand the mechanism.

Terminology provided above as a public service.

Why do you argue that unaccusatives are "just a special case" of ergative verbs? The label does serve some use to indicate that such verbs cannot be transitive. Granted, they are similar to/the same as the intransitive use of ergative verbs.

[Terra]

From the list below, you leave out unergative verbs. Even if my example of "barking" above doesn't hold, are you arguing that there is no nom/acc language anywhere that has an intransitive verb in which the subject has agent-like semantics?

No. I didn't want to include the "die" example in the first place, but did because no doubt somebody would ask about it, but couldn't think of something analagous for "unergative".

Saying that the noun w/ ergative case is an object in an abs/erg language is stretching the definition of "object" pretty far. If you mean "any verb argument that is not the subject", then yes. Even then, this would only hold true for syntactically ergative languages. A syntactically accusative language would probably consider the ergative noun to be the subject of a transitive sentence.

Yeah, my use of subject/object may be strange here. It seems better than using "first argument" and "second argument". If there's already some established terminology for what I used them for, then please enlighten me.

This does bring up an interesting point: would an antipassive voice ever be used with such a verb, since the intransitive use in this case would match the typical use of an antipassive?

Of course it would. English has "The window broke." and "The window was broken." after all.

Why do you argue that unaccusatives are "just a special case" of ergative verbs?

Warning: I have a feeling that you won't like the explanation.

The reason is because determining which argument is the agent/patient seems more fundamental/basic (and can be determined by a single argument (what I termed "subject") and the alignment of the language) than whether it can take a second argument. The latter seems like just an afterthought. Besides, why stop at two? "give" can take three, so what's the special word for it? (This doesn't mean that I don't wish that dictionaries would list the number of arguments that a verb can take, because I do.)

[Ser]

Oh gawd deer lawd, using English verbs in the passive voice to gloss verbs in an ergative-absolutive language. I had never seen that anyway... my brain hurts. I admit it works though (EDIT: even though glossing the agent argument of erg-abs languages with a prepositional phrase that works as an adjunct in English really threw me off — I don't think it's easy to catch that "by the dog" is supposed to be the subject of "was eaten" in "1) By the Dog.ABS was eaten (something/somebody)".).

[Terra]

Oh gawd deer lawd, using English verbs in the passive voice to gloss verbs in an ergative-absolutive language. I had never seen that anyway... my brain hurts. I admit it works though

I was trying to keep the SVO pattern to show how the patterns mirror eachother.

I don't think it's easy to catch that "by the dog" is supposed to be the subject of "was eaten" in "1) By the Dog.ABS was eaten (something/somebody)".).

Indeed. And if you were coming from an abs/erg language, you would say the same thing about "-- The window.NOM broke (because of something/somebody).". To mirror the same situation from abs/erg to nom/acc, Latin neuters end in -um in both the nominative and the accusative (think of -um being the absolutive, and look at type 2 verbs), which some say is because PIE was ergative at some point.

[Ser]

I don't think it's easy to catch that "by the dog" is supposed to be the subject of "was eaten" in "1) By the Dog.ABS was eaten (something/somebody)".).

Indeed. And if you were coming from an abs/erg language, you would say the same thing about "-- The window.NOM broke (because of something/somebody).".

No, I didn't say I found it confusing because it's just Different, coming from languages that are mostly nom-acc.

I said it's because you're using what is normally an adjunct (something that can be taken off and leave a sentence that is still grammatical) to gloss an argument here, in fact, the subject, of all things—all due to using English passive verbs to gloss verbs in an erg-abs language. It's strange to see "by the dog" as a subject of a verb, also such verbs in an erg-abs language would not be in something relatively marked like the passive voice.

[Terra]

No, I didn't say I found it confusing because it's just Different, coming from languages that are mostly nom-acc.

I said it's because you're using what is normally an adjunct (something that can be taken off and leave a sentence that is still grammatical) to gloss an argument here, in fact, the subject, of all things—all due to using English passive verbs to gloss verbs in an erg-abs language. It's strange to see "by the dog" as a subject of a verb, also such verbs in an erg-abs language would not be in something relatively marked like the passive voice.

What makes you think that if we were speaking in a abs-erg language, that I wouldn't gloss nom-acc verbs with the anti-passive and the subject with an adjunct in that certain case?

[Ser]

What makes you think I wouldn't point this out anyway?

[Terra]

What makes you think I wouldn't point this out anyway?

Doesn't that prove my point?

Anyways, Matsu, has all this answered your question? Is there anything that you don't understand?

[Ser]

What makes you think I wouldn't point this out anyway?

Doesn't that prove my point?

As I said above, you analysis works, I don't deny it. I just find it unintuitive because... c'mon I won't repeat it a third time.

Anyways, Matsu, has all this answered your question? Is there anything that you don't understand?

Actually, he hasn't logged in since March 11th 4:34 a.m. Vancouver time, and judging by the fact that the first response in this thread, finlay's, was posted at 8:49 a.m...

...In all likelihood he hasn't read a single response in this thread.

[matsu]

What makes you think I wouldn't point this out anyway?

Doesn't that prove my point?

As I said above, you analysis works, I don't deny it. I just find it unintuitive because... c'mon I won't repeat it a third time.

Anyways, Matsu, has all this answered your question? Is there anything that you don't understand?

Actually, he hasn't logged in since March 11th 4:34 a.m. Vancouver time, and judging by the fact that the first response in this thread, finlay's, was posted at 8:49 a.m...

...In all likelihood he hasn't read a single response in this thread.

I have! I have this post bookmarked on my phone. But work for me means long hours and I don't have much time to myself other than sleep. I'm about halfway through the linked paper on ergativity that was posted several posts up. I read it when I'm commuting.

I don't want you to think you have expended your energy to try and help me understand for nought.

[Vardelm]

I have! I have this post bookmarked on my phone. But work for me means long hours and I don't have much time to myself other than sleep. I'm about halfway through the linked paper on ergativity that was posted several posts up. I read it when I'm commuting.

Glad you're getting something out of it!

[matsu]

Ergative/Absolutive languages have two types of verbs:
1) When the verb takes an object, the subject is the patient and the object is the agent. However, when the verb lacks an object, the subject is the *agent*.
-- The apple.ABS was eaten by the dog.ERG.
-- By the dog.ABS was eaten (something/somebody).
2) When the verb takes an object, the subject is the patient and the object is the agent. When the verb lacks an object, the object/agent is implied: The subject is still the patient.
-- The window.ABS was broken by the man.ERG.
-- The window.ABS was broken (by something/someone).
3) The verb never takes an object, and the subject is the patient. (Again, this is really just a special case of type 2.)
-- The man.ABS was killed (by something/somebody).

I have a question about this. In the first type of verb, why does the agent switch from being ergative to absolutive? I understand the nominative-accusative equivalent example with "the window broke", but I never really liked that sort of construction in English because of the way English handles unaccusative verbs; the thematic role doesn't change, so should case change?

So, if a language marked "the dog ate []" as ABS, and also marked "the window was broken" as ABS, that's just essentially like the nominative-accusative languages mixing unergativity and unaccusativity into the same intransitive verb syntax structure? Aren't there examples of ABS-ERG languages where "the dog ate []" would preserve "dog" as ergative?

[matsu]

What makes you think I wouldn't point this out anyway?

Doesn't that prove my point?

Anyways, Matsu, has all this answered your question? Is there anything that you don't understand?

I'm going to have to expand my question to thematic roles as well, because it seems like an important part of understanding the distinction. On the Wikipedia article, there are examples of Basque sentences:

"Gizona etorri da."
man-ABS has arrived
'The man has arrived.'

"Gizonak mutila ikusi du."
man-ERG boy-ABS saw.
'The man saw the boy.'

I understand the second sentence; the man is the agent, so he is ergative, and the boy is the patient, so he is the absolutive. But when it comes to verbs of motion, like "arrive", why is the man in the absolutive? Surely he is probably responsible for his own motion? And from what I can tell, this seems to manifest itself in I-E as well, in "to be"/"to have" distinctions between unaccusative/unergative verbs, e.g. French. Is there a thematic role involved in this, different from agent?

Away from thematic roles explicitly, how do absolutive-ergative languages handle datives/indirect objects? Do some take different approaches from others? The McGregor paper doesn't satisfactorily address this for me.

[Vardelm]

In the first type of verb, why does the agent switch from being ergative to absolutive? ...... the thematic role doesn't change, so should case change?
----------------------
I'm going to have to expand my question to thematic roles as well, because it seems like an important part of understanding the distinction.

Yes, thematic roles are important, but I think you're overestimating their importance. There are syntactic/grammatical considerations as well as semantic ones, specifically the notions of "subject" and "object". These are not the same as "agent" & "patient"!

In the example of the dog & apple, "dog" changes noun case because it is the subject in the intransitive sentence, and the language in question calls for that subject to be in absolutive. It changes to ergative in the transitive sentence because the language requires the more agentive argument to be ergative. These are syntax issues operating at a different level than just thematic role. The syntactic notions of "subject" and "object" take thematic role into account, but there are other factors as well.

So, if a language marked "the dog ate []" as ABS, and also marked "the window was broken" as ABS, that's just essentially like the nominative-accusative languages mixing unergativity and unaccusativity into the same intransitive verb syntax structure?

Yes, it's a mixture of unergative & unaccusative verbs, just like happens in accusative languages. No language is going to be 100% ergative or accusative.

Aren't there examples of ABS-ERG languages where "the dog ate []" would preserve "dog" as ergative?
----------------------
I understand the second sentence; the man is the agent, so he is ergative, and the boy is the patient, so he is the absolutive. But when it comes to verbs of motion, like "arrive", why is the man in the absolutive? Surely he is probably responsible for his own motion?

I believe there are languages where "the dog ate" would have "dog" in the ergative case, although I can't name one off the top of my head. For verbs of motion (and other types), I think you're touching on the concept of volition, and there are definitely languages where intransitive subject could be either absolutive or ergative. A particular verb might call for its subject to specifically be in absolutive or ergative, or the speaker may have a choice depending on whether the subject acted intentionally or unintentionally.

These are examples of active-stative languages, not pure ergativity. Again, there are no languages that display pristine ergativity, so there will be some mix of other features included. It may be split-ergativity, where the language displays accusative behavior for certain tense/aspects, or person, etc. There could be a large number of accusative and/or unergative verbs. The language could be only morphologically/semantically ergative but have accusative syntax. Intransitives might display active-stative behavior.

On that note, accusative languages could display active-stative behavior as well, with agent-like verbs using the nominative, and patient-like verbs using the accusative as subject. In a language that does have active-stative features in intransitive sentences, the agentive or patientive argument of its transitive sentences being considered as "marked" will determine whether the language is "ergative with active-stative features" or "accusative with active-stative features".

It's because of all these ways in which languages can drift away from being purely accusative or ergative that I argue people should focus on what constitutes "pure ergativity", and then understand what features of pure ergativity the language has and what features it incorporates from other systems. I think you are focusing a lot on thematic roles and getting confused because there are languages that have some or many ergative features, but are not purely ergative, so they handle some of these issues (thematic role, volition, etc.) differently than a purely ergative language.

Away from thematic roles explicitly, how do absolutive-ergative languages handle datives/indirect objects? Do some take different approaches from others? The McGregor paper doesn't satisfactorily address this for me.

Same as accusative languages.

----------------------

Perhaps I can illustrate some of these topics with my own conlang. I consider it to be mostly ergative since it has ergative morphology & syntax. However, it does have some active-stative features in intransitive sentences. The word order is intransitive subject-verb, and transitive patient-verb-agent.

1) man.ABS die.INTR-PERF = "The man died."

2) man.ERG die.INTR-PERF = "The man killed himself / commited suicide."

3) man.ABS die.TRAN-PERF = "The man was killed (probably by someone/something else)."

4) man.ABS die.TRAN-PERF bear.ERG = "The man was killed by the bear."

5) ball.ABS give.TRAN-PERF girl.DATIVE boy.ERG = "The ball was given to the girl by the boy." -OR- "The boy gave the ball to the girl."


In the intransitive sentences #1 and #2, the only difference is whether "man" is marked in absolutive or ergative. The semantic difference is one of volition. It's interesting to note here that the man still has patientive semantics; he's the one that dies after all! So why am I using the ergative? Because the language has active-stative features, and uses the ergative to indicate volition as well as agentivity.

In sentence #3, the verb is now in a transitive form (.TRAN), and in the language this indicates that there is something, probably other than the man himself, which caused him to die. The thematic role of that unmentioned argument would be a cause or agent.

Sentence #4 specifically states that the direct agent of the man's death is a bear. A cause or force thematic role could also be included here, because the ergative case does not ONLY signify an agent thematic role.

Also, note that the subjects of sentences #1 and #2 are at the start of the sentence, regardless of whether they are absolutive or ergative. This makes the language have "subject-verb" order for intransitives. In sentence #4, the patient is the first word. Because the subject of intransitives and the patient of transitives occupy the same position, the language has ergative word order (a feature of syntax). If the language had the ergative noun in sentence #2 following the verb (verb-subject), then the language would have active-stative word order/syntax, as well as active-stative morphology & semantics. The way my language works makes it have active-stative morphology/semantics, but not active-stative syntax.

Finally, sentence #5 simply shows the use of a direct and indirect object. The indirect object (girl.DATIVE) could be placed anywhere depending on the language. In mine, it happens to be required to follow right after the verb. I might allow it to also be in other positions depending on certain requirements, but I haven't figured that out quite yet.

[Cedh]

On that note, accusative languages could display active-stative behavior as well, with agent-like verbs using the nominative, and patient-like verbs using the accusative as subject.

You don't actually have to look very far to find examples of this. Various Germanic languages have "quirky case subjects" in some intransitive verbs of low volition. Most commonly it's the dative case that is used here, but the accusative also occurs.

German:

Mir ist kalt.
me[DAT] is cold
"I'm cold."

Mich friert. (slightly archaic, but still grammatical)
me[ACC] freezes.
"I'm freezing."

Icelandic:

Mér kólnar.
me[DAT] is_getting_cold
"I'm getting cold."

Mig kelur.
me[ACC] is_freezing
"I'm freezing."

Even English has some remnants of this, for instance the fossilized contraction methinks (cf. German mich dünkt, which is also slightly archaic but still grammatical in formal registers).

[matsu]

Thank you all for your input. Terra, Valdelm, thank you two especially. I'd never understood the ergative-absolutive alignment, and for the first time I have a good grasp of it. I was marking transitivity and unaccusative/unergative types of intransitivity separately because I couldn't resolve thematic role issues, but I think ergative-absolutive will allow me to do so. I'm also really fond of the idea of volition arguments, and volition affecting ergative or absolutive role in language.

Thanks!