Understanding ergativity

[Gaspard]

Hey folks,

I’m trying to build an ergative conlang, and I want to make sure I understand the concept before continuing, so here is what I have so far:

intransitive verb:
(1) be.happy[-3.SG] man-ABS
The man is happy.

transitive verb:
(2) love[-3.SG] man-ERG mountain-ABS-PL
The man loves the mountains.

The language is VSO, and the second example shows that the transitive verb agrees with the ergative subject. Is it plausible to do this? Would it be plausible to not mark subject and object at all on the verb? [I put it in brackets, since I don’t know yet if I want to mark the subject on the verb.]
(I’ve read somewhere that in ergative languages the verb tends to always agree with the absolutive part of the sentence. But I don’t see why in that case one doesn’t simply call the absolutive nominative, the ergative whatever, and the language a nom.-acc. language…)

Now antipassive. As far as I understand it, the ergative subject of a transitive verb shifts to absolutive and the absolutive object shifts somewhere else (f. ex. oblique), like so:

(3) love-AP[-3.SG] man-ABS mountain-OBL-PL
The man loves the mountains.

I marked the antipassive on the verb, like a passive would be marked on the verb as well. The translation is the same, as English can’t simulate an antipassive construction.

What would you need the antipassive for? Well, for sentences like this:

(4) be.happy[-3.SG] man-ABS when see-AP-3.SG mountain-OBL-PL
The man is happy when he sees the mountains.

Since both subjects are the same they have to be in the same case (here absolutive), so the sentences can be connected. Otherwise the mountain would be happy…

Last thing: if you want to drop the ergative subject from a sentence like this:

(5) see-[3.SG] man-ERG mountain-ABS-PL
The man sees the mountains.

you just do so – there is no need for a passive voice. However I fear this doesn’t make too much sense if I mark the subject on the verb, since then I am not really dropping anything:

(6) see-[3.SG] mountain-ABS-PL
He sees the mountains. / The mountains are seen.

Is this the way an ergative language would work? Am I forgetting anything important?
Thanks for any feedback!

Greetz,
Gaspard

[Vardelm]

I’m trying to build an ergative conlang, and I want to make sure I understand the concept before continuing,

Take a look at the following linked PDF. I always point people to this when they have questions about ergativity. I've never seen a better breakdown of the different types of ergativity. It helps make it clear where languages like Basque that are morphologically ergative but syntactically accusative sit in the accusative <> ergative spectrum.

Typology of Ergativity - William McGregor

intransitive verb:
(1) be.happy[-3.SG] man-ABS
The man is happy.

Yep, this is OK.

transitive verb:
(2) love[-3.SG] man-ERG mountain-ABS-PL
The man loves the mountains.

The language is VSO, and the second example shows that the transitive verb agrees with the ergative subject. Is it plausible to do this?

Yes, it's quite plausible, but making the verb agree with the ergative subject means that the language is morphologically ergative and syntactically accusative (as far as verb-noun agreement goes). AFAIK, more languages that have ergative features work like this than those that have syntactic ergativity (such as making the verb agree with the object).

Would it be plausible to not mark subject and object at all on the verb? [I put it in brackets, since I don’t know yet if I want to mark the subject on the verb.]

I think so, yes. There are languages that don't mark person on the verb at all, and ergative languages are no different than accusative languages in terms of other features.

(I’ve read somewhere that in ergative languages the verb tends to always agree with the absolutive part of the sentence. But I don’t see why in that case one doesn’t simply call the absolutive nominative, the ergative whatever, and the language a nom.-acc. language…)

If the language is purely ergative or close to it (it has morphological, syntactic, AND semantic ergativity, according to McGregor's terms), then yes, verbs would probably agree more with the absolutive. But again, most languages with ergative features tend to have a number of accusative features as well. It makes discussion of ergativity messy & confusing, & it's why McGregor's paper is so valuable. Rather than talking about languages being entirely ergative or accusative, it breaks down different areas of ergativity & accusativity so that we can look at it in a slightly more granular manner that is still organized well.

As to why it wouldn't simply be an accusative language at that point gets into the semantics of which argument is marked. I'm somewhat time-limited at the moment, but I can explain more later if needed.

Now antipassive. As far as I understand it, the ergative subject of a transitive verb shifts to absolutive and the absolutive object shifts somewhere else (f. ex. oblique), like so:

(3) love-AP[-3.SG] man-ABS mountain-OBL-PL
The man loves the mountains.

I marked the antipassive on the verb, like a passive would be marked on the verb as well. The translation is the same, as English can’t simulate an antipassive construction.

This looks OK as well.

What would you need the antipassive for? Well, for sentences like this:

(4) be.happy[-3.SG] man-ABS when see-AP-3.SG mountain-OBL-PL
The man is happy when he sees the mountains.

Since both subjects are the same they have to be in the same case (here absolutive), so the sentences can be connected. Otherwise the mountain would be happy…

Yep, I think ergative languages often require an antipassive in order to form relative clauses & such.

Also, the antipassive can have some subtle changes in meaning compared to a transitive construction, which can include definiteness or telicity. So your example above might mean "The man loves mountains" rather than "THE mountains". Exactly what the antipassive means or is used for varies from language to language, so if you are only using it for coordination, that's probably fine.

Last thing: if you want to drop the ergative subject from a sentence like this:

(5) see-[3.SG] man-ERG mountain-ABS-PL
The man sees the mountains.

you just do so – there is no need for a passive voice. However I fear this doesn’t make too much sense if I mark the subject on the verb, since then I am not really dropping anything:

(6) see-[3.SG] mountain-ABS-PL
He sees the mountains. / The mountains are seen.

Yes, that's plausible as well. Again, ergative languages aren't any different from others, so some will allow arguments to be dropped, and others won't.

Also, you could still have a passive voice (if you want), but like the antipassive it might carry slight changes in meaning vs. your example #6.

Is this the way an ergative language would work? Am I forgetting anything important?

It's the way SOME ergative languages work. Ergative languages are not all the same. They use ergativity in different ways, in different parts of the language. Take the time to really go through that PDF and understand it. It will really help you clarify why everything works the way it does, and then design your language to work how you want it to.

Start thinking about different types of verbs. The examples you used are emotions & perceptions, and often those work differently than other verbs in some languages. A couple good ones to think about:

The bread bakes.
The chef bakes.
The chef bakes the bread.

The hunter shot.
The bear was shot.
The hunter shot the bear.

In English, "bake" is an ergative verb, while "shoot" is accusative. So yeah, eventually you'll need to get into ergative, accusative, unergative, and unaccusative verbs. FUN!!!!! I would suggest to read the PDF, think about the above examples & other ones before you start reading too much about these types of verbs. You'll get yourself into a quagmire fast if you're not prepared for it.

Good luck, have fun, and feel free to ask more questions.

[Imralu]

Hey folks,

Hi Berliner!

The language is VSO, and the second example shows that the transitive verb agrees with the ergative subject. Is it plausible to do this? Would it be plausible to not mark subject and object at all on the verb? [I put it in brackets, since I don’t know yet if I want to mark the subject on the verb.]
(I’ve read somewhere that in ergative languages the verb tends to always agree with the absolutive part of the sentence. But I don’t see why in that case one doesn’t simply call the absolutive nominative, the ergative whatever, and the language a nom.-acc. language…)

You might be interested to read a little bit about Tongan. Polynesian languages are VSO (or, I think a couple might be VOS) and have no personal agreement on their verbs. Many of them, such as Maori are basically NOM-ACC, but others such as Tongan are basically ERG-ABS. It's a bit of a blurry line though. Essentially it depends how normal it is for the verbs to be in the passive, whether that is the marked form or not.

Example from Maori.

Active:

• Kua hanga ngā tohunga i te marae.
  PRF build DEF.PL expert ACC DEF.SG marae.
  The experts have built the marae.

Passive:

• Kua hangaia te marae e ngā tohunga.
  PRF build-PASS DEF.SG marae AGENT DEF.PL expert
  The marae has been built by the experts.

Maori tends to use the passive a great deal more than English. If we regard hangaia as the normal verb form, then it basically becomes ERG-ABS.

• Kua hangaia te marae e ngā tohunga.
  ?PRF build DEF.SG marae ERG DEF.PL expert
  ?The experts built the marae.

Of course, this is possibly not the correct interpretation of the Maori sentence as the passive verb form appears to be at least slightly more marked and also probably places pragmatic emphasis differently in the sentence.

From what I've seen of Niuean, which is supposedly ERG-ABS, the verbs look like the passive form of their Maori cognates (Maori kite-a 'see-PASS', Niuean kitia 'see'). Incidentally, tohunga is cognate to the well-known Hawai'ian word kahuna.

My conlang, Ngolu is also VSO. It's generally pretty clearly nominative accusative, but sometimes it seems that you could analyse it a bit differently, with the causal case as the ergative case and the nominative as the absolutive.

• kaxina xu xuomaki
  shatter.INTR NOM.the window
  The window shatters.
  
  kaxina teui xu xuomaki
  shatter.INTR CAU.2s NOM.the window
  You shatter the window.
  
  zaxa nu
  be.sick NOM.1s
  I'm sick.
  
  zaxa nu teui
  be.sick NOM.1s CAU.2s
  ≈ sicken ABS.1s ERG.2s
  You made me sick. / I'm sick because of you.

However, there are also causative verbs derivable from both of the above and then the pattern is definitely NOM-ACC although it's a bit weird because any and all arguments can be omitted.

• kaxua vu
  shatter.TR NOM.2s
  You shatter something.
  
  kaxua vu xi xuomaki
  shatter.TR NOM.2s ACC.the window
  You shatter the window.
  
  kaxua xi xuomaki
  shatter.TR ACC.the window
  Someone shatters the window. / The window is shattered.
  
  zaxua vu
  sicken NOM.2s
  You make someone sick. [non-metaphorically]
  
  zaxua vu ni
  sicken NOM.2s ACC.1s
  You make me sick. [non-metaphorically]
  
  zaxua ni
  sicken ACC.1s
  Someone makes me sick. / I am sickened. [non-metaphorically]

[Vardelm]

You might be interested to read a little bit about Tongan. Polynesian languages are VSO (or, I think a couple might be VOS) and have no personal agreement on their verbs. Many of them, such as Maori are basically NOM-ACC, but others such as Tongan are basically ERG-ABS. It's a bit of a blurry line though. Essentially it depends how normal it is for the verbs to be in the passive, whether that is the marked form or not.

I would sound 1 note of caution here since these are Austronesian languages. I don't know enough about Maori & Tongan to say whether they are examples of "Austronesian" alignment per se, but from what I understand the Austronesian languages tend to have a lot of ergative & accusative features intermingled. In my view, this makes it important to understand what a "purely ergative" language would look like, even though such an animal doesn't truly exist. It's just a measuring stick, if that makes sense.

Maori tends to use the passive a great deal more than English. If we regard hangaia as the normal verb form, then it basically becomes ERG-ABS.

Not surprising, since the "passive" in Malay is, from what I've read, more commonly used than the "active". The 2 voices have about the same markedness.

My conlang, Ngolu is also VSO. It's generally pretty clearly nominative accusative, but sometimes it seems that you could analyse it a bit differently, with the causal case as the ergative case and the nominative as the absolutive.

zaxa nu
be.sick NOM.1s
I'm sick.

zaxa nu teui
be.sick NOM.1s CAU.2s
≈ sicken ABS.1s ERG.2s
You made me sick. / I'm sick because of you.
[/list]

I think that's a fair analysis. In many ergative languages, the ergative case is actually the same as another case, such as the genitive, instrumental, or ablative. This might confuse things a bit, since there could be debate whether you are dealing with a transitive sentence w/ 2 core arguments, or an intransitive sentence with 1 core and 1 oblique.

[Curlyjimsam]

This thread is quite old now, but a couple of hopefully useful comments.

(1) For some reason ergative languages with orders other than SOV are pretty rare (I don't think anybody really knows why). This isn't to say a VSO ergative language couldn't happen, and there are purported instances of such, but such a thing would be atypical.

(2) Similarly, for some reason if a language has ergative-absolutive agreement, then it will only have agreement the ergative argument if it also has agreement with the absolutive. Again there are a couple of apparent exceptions but agreement with only the ergative argument is at best very rare.

[Having agreement with only the absolutive argument does not make that an nominative-accusative alignment: absolutive agreement targets the intransitive argument and the transitive object (S/P), whereas nominative agreement targets the intransitive argument and the transitive subject (S/A). But it is quite common for a language to have agreement which follows a nominative-accusative pattern and case which follows an ergative-absolutive pattern.]