Nom/Acc vs. Erg/Abs

If you're new to these arts, this is the place to ask "stupid" questions and get directions!
Post Reply
User avatar
Sodomor
cuneiform
cuneiform
Posts: 82
Joined: 18 Mar 2011 13:18
Location: St. Paul, MN

Nom/Acc vs. Erg/Abs

Post by Sodomor »

Ok, I understand the differences between nominative/ accusative pangs and ergative/absolutive langs, but I was wonder what the "side effects" are.

Someone on here mentioned that nom/acc languages favor the passive and erg/abs languages favor the anti passive, but I don't understand why.

And are there any other differences like that?
User avatar
eldin raigmore
korean
korean
Posts: 6352
Joined: 14 Aug 2010 19:38
Location: SouthEast Michigan

Re: Nom/Acc vs. Erg/Abs

Post by eldin raigmore »

Sodomor wrote:... I understand the differences between nominative/ accusative pangs and ergative/absolutive langs, ...
I've never really thought of it as painful.

But, all that those two morphosyntactic-alignment types differ in, is, which of the two participants in transitive clauses is treated just like the subject of intransitive clauses, and which of them is instead treated in some different way.

If you want to see what correlates with alignment-type, see http://wals.info/feature/98A and its associated chapter; also, combine feature 98A with any other feature you want to see about.

For instance, http://wals.info/feature/combined/98A/99A,
or http://wals.info/feature/combined/98A/121A,
or http://wals.info/feature/combined/81A/98A,
or http://wals.info/feature/combined/102A/98A.

Someone on here mentioned that nom/acc languages favor the passive and erg/abs languages favor the anti passive, but I don't understand why.
In an acc/nom transitive clause, the Agent is usually the Subject; the Agent is usually the Nominative argument.
If you want to change that you have to use a voice-transformation that promotes the Patient to Subject.
Any such voice-transformation is called Passivization; any clause in which some voice-operation has happened to a transitive verb to allow its Subject to be its Patient, is a "passive" clause.
OTOH, if you want to drop the Patient and not mention it, that doesn't involve promoting the Patient (which was the Object) nor demoting the Agent (which was and still is the Subject).
Any voice-operation which just drops the patient is "De-Patientive", not "Anti-Passive".
So acc/nom languages don't need "Anti-Passives".


In an erg/abs transitive clause, the Patient is usually the Absolutive (sometimes aka Nominative) argument.
If you want to change that and make the Agent be the Absolutive argument, you have to use a voice-transformation that promotes the Agent from Ergative to Absolutive.
Any such voice-transformation is called "Anti-Passive".
If you want to drop the Patient and mention only the Agent, then, since the Agent is going to be the only participant of the new clause, it will have to be Absolutive; that means it will have to be promoted out of Ergative into Absolutive. That's still "Anti-Passive".
OTOH, if you want to just drop the Agent and not mention it, that doesn't involve promoting the Agent (which was the Ergative) nor demoting the Patient (which was and still is the Absolutive).
Any voice-operation which just drops the Agent is "De-Agentive", not "Passive".
So erg/abs languages don't need "Passives".
User avatar
Sankon
sinic
sinic
Posts: 357
Joined: 12 Aug 2010 02:06
Location: At the computer

Re: Nom/Acc vs. Erg/Abs

Post by Sankon »

Read this. It's an overview of ergativity and its ramifications, and it greatly helps with understanding how and why ergative alignment happens.
User avatar
eldin raigmore
korean
korean
Posts: 6352
Joined: 14 Aug 2010 19:38
Location: SouthEast Michigan

Re: Nom/Acc vs. Erg/Abs

Post by eldin raigmore »

Sankon wrote:Read this. It's an overview of ergativity and its ramifications, and it greatly helps with understanding how and why ergative alignment happens.
Do understand, though, that their terminology can be confusing. (Not only theirs.)

When they say "subject" they actually mean "agent"; when they say "object" they actually mean "patient".

"Agent" and "patient" are semantic roles. In a prototypically transitive clause, the agent acts directly and wilfully upon the patient, and the patient is physically and visibly affected by the agent.
User avatar
Micamo
MVP
MVP
Posts: 5671
Joined: 05 Sep 2010 19:48
Contact:

Re: Nom/Acc vs. Erg/Abs

Post by Micamo »

Sankon wrote:Read this. It's an overview of ergativity and its ramifications, and it greatly helps with understanding how and why ergative alignment happens.
You know, ergative switch-referent markers are a fantastic idea...
My pronouns are <xe> [ziː] / <xym> [zɪm] / <xys> [zɪz]

My shitty twitter
User avatar
Chagen
runic
runic
Posts: 3338
Joined: 03 Sep 2011 05:14
Location: Texas

Re: Nom/Acc vs. Erg/Abs

Post by Chagen »

I still find it strange that Tripartite is so damn rare.

Föro is perfectly tripartite under all situations, not a single split-ergativity affair here.

Interestingly, one of my other conlangs, Cryset, is heavily analytical, with grammar heavily inspired by English....but it shows a Tripartite system in Pronouns (like, English's nominative/objective split in pronouns, just with a tripartite system.
Nūdenku waga honji ma naku honyasi ne ika-ika ichamase!
female-appearance=despite boy-voice=PAT hold boy-youth=TOP very be.cute-3PL
Honyasi zō honyasi ma naidasu.
boy-youth=AGT boy-youth=PAT love.romantically-3S
User avatar
Sankon
sinic
sinic
Posts: 357
Joined: 12 Aug 2010 02:06
Location: At the computer

Re: Nom/Acc vs. Erg/Abs

Post by Sankon »

Chagen wrote:Föro is perfectly tripartite under all situations, not a single split-ergativity affair here.
That seems really unlikely. All languages have at least some features of both ergativity and accusativity in them, even English. Just look at English's ergative verbs: 'I walk" vs. "I walk my dog", "It's cooking" vs. "I'm cooking it", "It burst" vs. "I burst your bubble".

Don't ask me why, most likely it has something to do with animacy, but no languages (to my knowledge) are completely, 100%, entirely accusative/ergative/tripartite/whatever.
User avatar
Chagen
runic
runic
Posts: 3338
Joined: 03 Sep 2011 05:14
Location: Texas

Re: Nom/Acc vs. Erg/Abs

Post by Chagen »

All languages have at least some features of both ergativity and accusativity in them
Uh, isn't the entire point of a Tripartite system that it mixes both ergativity and accusativity, as you said?
Nūdenku waga honji ma naku honyasi ne ika-ika ichamase!
female-appearance=despite boy-voice=PAT hold boy-youth=TOP very be.cute-3PL
Honyasi zō honyasi ma naidasu.
boy-youth=AGT boy-youth=PAT love.romantically-3S
User avatar
Micamo
MVP
MVP
Posts: 5671
Joined: 05 Sep 2010 19:48
Contact:

Re: Nom/Acc vs. Erg/Abs

Post by Micamo »

I've never read any attempt at a formal analysis of tripartite phenomena, so forgive me if I'm pulling things out of my ass here. But I can understand the idea that tripartism is something wholly separate from either accusativity or ergativity. Accusativity is the conflation of I and A, while Ergativity is the conflation of I and P. In a tripartite system, none of the three are conflated. I don't think tripartism is as simple as "accusative and ergative mixed together."
My pronouns are <xe> [ziː] / <xym> [zɪm] / <xys> [zɪz]

My shitty twitter
User avatar
eldin raigmore
korean
korean
Posts: 6352
Joined: 14 Aug 2010 19:38
Location: SouthEast Michigan

Re: Nom/Acc vs. Erg/Abs

Post by eldin raigmore »

Micamo wrote:You know, ergative switch-referent markers are a fantastic idea...
My conlang Adpihi's switch-reference system does that.
For any Consecutive (i.e. non-Initial) clause (the Initial clause is the anchor clause), the first (or only) morpheme of the switch-reference marker shows one of five values:
  1. The intransitive Subject or the Actor (think "Agent") of the marked clause is the same as the S or the A of the referred-to clause;
  2. Not (1), but the S or the A of the marked clause is the same as the Undergoer (think "Patient") or Extended-Core Term (think "Dative") of the referred-to clause;
  3. Neither of the above, but the S or the A of the marked clause properly contains, or is properly contained in, the S or the A of the referred-to clause;
  4. None of the above, but the S or the A of the marked clause properly contains, or is properly contained in, the U or the E of the referred-to clause;
  5. None of the above.
But, if the marked clause is transitive (has both an Actor and an Undergoer), then the switch-reference marker has a second morpheme that tells about its U;
  1. The Undergoer of the marked clause is the same as the S or the U of the referred-to clause;
  2. Not (1), but the U of the marked clause is the same as the A or E of the referred-to clause;
  3. Neither of the above, but the U of the marked clause properly contains, or is properly contained in, the S or the U of the referred-to clause;
  4. None of the above, but the U of the marked clause properly contains, or is properly contained in, the A or the E of the referred-to clause;
  5. None of the above.
So it tracks Same-or-Different-Agent-(or-Subject) and also Same-or-Different-Patient-(or-Object).

If I changed the description of the first morpheme to say Undergoer wherever it now says Actor, then also changed the description of the second morpheme to say Actor wherever it now says Undergoer, that would be an ergative switch-reference system, right?

@Chagen:
Accusative/Nominative systems, and Ergative/Absolutive systems, both use just two "cases" or treatments, to handle the three semantic roles of intransitive Subject, transitive Agent or Actor, and transitive Undergoer or Object or Patient.
That's economical.

Another system that uses just two treatments ("cases") to handle the three roles, is the Split-Intransitive or Split-S or Active/Stative system. In it, some intransitive Subjects (called SA) are treated like transitive Agents (call those "nominative"), but other intransitive Subjects (called SP) are treated like transitive Patients (call those "absolutive"). A subtype of Split-S is Fluid-S; in those, there are sometimes two intransitive clauses with the same verb and the same subject noun, but in one it's marked as if it were an Agent and in the other it's marked as if it were a Patient; and the two clauses mean slightly different things.

Tripartite, on the other hand, uses three different treatments ("cases") to handle the three roles; an ergative for Agents, an accusative for Patients, and a nominative/absolutive for intransitive Subjects. That's not economical.

Another system with three different treatments is the Split-Transitive system, aka "Austronesian-Phillippine System". In that system, a transitive clause will always have one participant marked (or, rather, unmarked) the same as an intransitive Subject (probably called "nominative" or "absolutive"), but which one that is will vary. There's a "case" or treatment that marks Agents (let's call it "ergative"); if one participant is marked as an Agent, the one that's unmarked must be the Patient. There's a third "case" or treatment that marks Patients (let's call it "accusative"); if one participant is marked as a Patient, the participant that isn't marked must be the Agent.
That, likewise, is not economical.

There's a system politely called "transitive alignment" but probably more informatively called "monster raving looney alignment", with just two "cases", in which the intransitve Subject is treated one way, and both the transitive Agent and the transitive Patient are treated the other way. That's a waste of a case-distinction, since the intransitive Subject will never occur in the same clause as either of the other two roles and so doesn't need to be distinguished from either of them by case, but the Agent and Patient will occur in the same clause and so will need to be distinguished from one another somehow.

Also, there's a difference between "morphological ergativity" and "syntactic ergativity". Languages that are both morphologically and syntactically ergative are rarer than those that are just morphologically ergative.

And morphological case-marking of the nouns and pronouns may be aligned differently from the morphological agreement-marking of the verbs, though often they're the same.


____________________________________________________________________________

Does any of that help?


EDIT: Should I have put this in the "Beginners' Corner" subforum, or moved it to another thread?
Last edited by eldin raigmore on 18 May 2012 21:31, edited 1 time in total.
User avatar
cybrxkhan
roman
roman
Posts: 1106
Joined: 25 Dec 2010 21:21
Contact:

Re: Nom/Acc vs. Erg/Abs

Post by cybrxkhan »

As a side question, were there any historical instances where a nom/acc natlang turned into erg/abs or vice versa? I don't mean when the natlang is erg/abs in a few specific situations but nom/acc otherwise (or vice versa), but when there is a clear shift.
I now have a blog. Witness the horror.

I think I think, therefore I think I am.
- Ambrose Bierce
User avatar
Micamo
MVP
MVP
Posts: 5671
Joined: 05 Sep 2010 19:48
Contact:

Re: Nom/Acc vs. Erg/Abs

Post by Micamo »

eldin raigmore wrote:My conlang Adpihi's switch-reference system does that.
For any Consecutive (i.e. non-Initial) clause (the Initial clause is the anchor clause), the first (or only) morpheme of the switch-reference marker shows one of five values:
  1. The intransitive Subject or the Actor (think "Agent") of the marked clause is the same as the S or the A of the referred-to clause;
  2. Not (1), but the S or the A of the marked clause is the same as the Undergoer (think "Patient") or Extended-Core Term (think "Dative") of the referred-to clause;
  3. Neither of the above, but the S or the A of the marked clause properly contains, or is properly contained in, the S or the A of the referred-to clause;
  4. None of the above, but the S or the A of the marked clause properly contains, or is properly contained in, the U or the E of the referred-to clause;
  5. None of the above.
But, if the marked clause is transitive (has both an Actor and an Undergoer), then the switch-reference marker has a second morpheme that tells about its U;
  1. The Undergoer of the marked clause is the same as the S or the U of the referred-to clause;
  2. Not (1), but the U of the marked clause is the same as the A or E of the referred-to clause;
  3. Neither of the above, but the U of the marked clause properly contains, or is properly contained in, the S or the U of the referred-to clause;
  4. None of the above, but the U of the marked clause properly contains, or is properly contained in, the A or the E of the referred-to clause;
  5. None of the above.
So it tracks Same-or-Different-Agent-(or-Subject) and also Same-or-Different-Patient-(or-Object).

If I changed the description of the first morpheme to say Undergoer wherever it now says Actor, then also changed the description of the second morpheme to say Actor wherever it now says Undergoer, that would be an ergative switch-reference system, right?
Hey Eldin, you should write a guide on the different options languages have in clause-chaining!
My pronouns are <xe> [ziː] / <xym> [zɪm] / <xys> [zɪz]

My shitty twitter
User avatar
eldin raigmore
korean
korean
Posts: 6352
Joined: 14 Aug 2010 19:38
Location: SouthEast Michigan

Re: Nom/Acc vs. Erg/Abs

Post by eldin raigmore »

cybrxkhan wrote:As a side question, were there any historical instances where a nom/acc natlang turned into erg/abs or vice versa? I don't mean when the natlang is erg/abs in a few specific situations but nom/acc otherwise (or vice versa), but when there is a clear shift.
ISTR I believe someone posted somewhere (here?) a link to a scholarly paper that "proved", as near as such things can be proven, how one of the Inuit or Yupik or other "Eskimo" languages, switched to erg/abs (from acc/nom IIRC).
But TTBOMK no-one knows of any such switch happening in history, that is, in historical times so that the switch was actually recorded. It seems doubtful; most major language-change happens where history can't see it.

But, find an erg/abs language whose subordinate clauses are mostly acc/nom; chances are that's a language in the process of shifting from acc/nom to erg/abs.
And/or, find an acc/nom language whose subordinate clauses are mostly erg/abs; chances are that's a language in the process of shifting from erg/abs to acc/nom.

AIUI, both directions of shifts are accepted diachronics for some extant natlangs. Sorry I don't remember which ones.
User avatar
cybrxkhan
roman
roman
Posts: 1106
Joined: 25 Dec 2010 21:21
Contact:

Re: Nom/Acc vs. Erg/Abs

Post by cybrxkhan »

eldin raigmore wrote:
cybrxkhan wrote:As a side question, were there any historical instances where a nom/acc natlang turned into erg/abs or vice versa? I don't mean when the natlang is erg/abs in a few specific situations but nom/acc otherwise (or vice versa), but when there is a clear shift.
ISTR I believe someone posted somewhere (here?) a link to a scholarly paper that "proved", as near as such things can be proven, how one of the Inuit or Yupik or other "Eskimo" languages, switched to erg/abs (from acc/nom IIRC).
But TTBOMK no-one knows of any such switch happening in history, that is, in historical times so that the switch was actually recorded. It seems doubtful; most major language-change happens where history can't see it.

But, find an erg/abs language whose subordinate clauses are mostly acc/nom; chances are that's a language in the process of shifting from acc/nom to erg/abs.
And/or, find an acc/nom language whose subordinate clauses are mostly erg/abs; chances are that's a language in the process of shifting from erg/abs to acc/nom.

AIUI, both directions of shifts are accepted diachronics for some extant natlangs. Sorry I don't remember which ones.
Thanks for the info. So it seems possible but very much unattested at this point, then.

I was thinking of having a erg/abs protolang that had a large chunk of nom/acc daughterlangs (potentially even the majority of them), my reason for asking. But I think I'll just save the erg/abs to a smaller language family on Arteran to enforce my discrimination towards them. [>:D]
I now have a blog. Witness the horror.

I think I think, therefore I think I am.
- Ambrose Bierce
User avatar
eldin raigmore
korean
korean
Posts: 6352
Joined: 14 Aug 2010 19:38
Location: SouthEast Michigan

Re: Nom/Acc vs. Erg/Abs

Post by eldin raigmore »

Micamo wrote:Hey Eldin, you should write a guide on the different options languages have in clause-chaining!
Look at the "spoilered" Preliminaries 1 and Preliminaries 2 in this post.

For a better guide you should look at Switch Reference and Universal Grammar: Proceedings of a symposium on switch reference and universal grammar, Winnipeg, May 1981: Edited by John Haiman and Pamela Munro: Typological Studies in Language 2

Also see this search.

And read the Wikipedia article; it might not be perfect but it's a great start.

Mostly switch-reference marks one of:
same-or-different-Subject (usually that means "same-or-different Agent")
same-or-different-Object (usually that means "same-or-different Patient")
same-or-different-Place
same-or-different-Time (but this may be no different than "relative tense" and "relative aspect").

By far the most common is "same-or-different-Subject".

"same-or-different" doesn't just mean "same-or-different", though.

For one thing: The SR marker can have values comparing the marked clause's Subject to other participants of the referred-to clause other than its Subject. For instance, it can have a "marked-clause's-Subject-is-same-as-referenced-clause's-Subject" value, and a "marked-clause's-Subject-is-same-as-referenced-clause's-Direct-Object" value, and a "marked-clause's-Subject-is-same-as-referenced-clause's-Indirect-Object" value, and a a "marked-clause's-Subject-is-not-the-same-as-any-of-referenced-clause's-core-arguments" value.

For another thing: The SR marker can have values besides just "same" and "different". Often, one direction or the other of proper containment is distinguished, when at least one of the concerned participants is a group (e.g. plural or dual). For instance, if we just compare the marked clause's Subject to the referenced clause's Subject, the language may have one of the following four systems:
  1. The SR has three values;
    1. one value means the marked clause's subject is either the same as or properly contains the referenced clause's subject;
    2. one value means the marked clause's subject is properly contained in the referenced clause's subject;
    3. one value means the two subjects are different, and neither of the two subjects contains the other.
  2. The SR has three values;
    1. one value means the marked clause's subject is either the same as or is properly contained in the referenced clause's subject;
    2. one value means the marked clause's subject properly contains the referenced clause's subject;
    3. one value means the two subjects are different, and neither of the two subjects contains the other.
  3. The SR has three values;
    1. one value means the marked clause's subject is the same as the referenced clause's subject;
    2. one value means the marked clause's subject properly contains the referenced clause's subject;
    3. one value means the two subjects are different, and the marked clause's subject does not contain the referenced clause's subject.
  4. The SR has three values;
    1. one value means the marked clause's subject is the same as the referenced clause's subject;
    2. one value means the marked clause's subject is properly contained in the referenced clause's subject;
    3. one value means the two subjects are different, and the marked clause's subject is not contained in the referenced clause's subject.


Furthermore, chaining can be either "backward" or "forward".
For instance, maybe the referenced clause is always the first clause in the chain;
or, maybe the referenced clause is always the clause immediately preceding the marked clause. Either of those would be "backwards chaining"; so would any other arrangement in which the referenced clause was always before the marked clause.
Or, maybe the referenced clause is always the last clause in the chain;
or, maybe the referenced clause is always the clause immediately following the marked clause. Either of those would be "forwards chaining"; so would any other arrangement in which the referenced clause was always after the marked clause.

In backwards-chaining languages, the first clause in the chain is its "anchor clause" and is often (or at least sometimes) the only finite clause (the only clause with a finite verb) in the chain. It occurs in an Initial or Independent mood or order; all the later clauses in the chain occur in a Consecutive, or Conjunctive, or Conjunct, mood or order.
In forwards-chaining languages, the last clause in the chain is its "anchor clause" and is often (or at least sometimes) the only finite clause (the only clause with a finite verb) in the chain. It occurs in a Final or Independent mood or order; all the earlier clauses in the chain occur in a Medial, or Conjunctive, or Conjunct, mood or order.

Forwards-chaining vs backwards-chaining is statistically correlated to OV word-order vs VO word-order, but I don't remember how exactly.

There are several languages whose single clauses seem to be thoroughly erg/abs, but whose clause-chaining and switch-reference system seems to be thoroughly acc/nom (for instance it tracks Agents rather than Patients).

If you look up switch-reference and clause-chaining languages you'll probably run into forwards-chaining languages before backwards-chaining ones.

Clause-chaining is a substitute for subordination. In a two-clause chain one of them is conjoined to the other in a way midway between co-ordination and sub-ordination. (In a longer chain there may be a web or tree, or at least a chain, of such not-quite-coordinate not-quite-subordinate conjoined relationships between the clauses.)

If the language changes, for instance its word-order changes or for some reason speakers want relative clauses to modify each of three or more participants in a main clause, the clause-chaining may evolve into true subordination. (In clause-chaining a "relative clause" can modify only the first or the last participant in a "mainer" clause.)

Some of the markers, for instance relative pronouns, that show that some participant of a subordinate clause is the same as some participant of its matrix clause or of the main clause, are sometimes called "switch-reference markers". One way they could arise is by the type of language change mentioned in the preceding paragraph.

In languages in which the head-noun (the noun the RC modifies, the shared participant with the MC) of a relative clause must be its subject, these markers would show which of the MC's participants the RC's subject is the same as. But if the RC's head could be an object, it might show other things as well or instead .
Post Reply