Passive voice in Ergative languages

[Grimalkin]

So, many ergative languages don't have a passive voice, choosing instead to have an antipassive. Now, what I'm wondering is how exactly do languages with an antipassive but no passive voice demote the agent or promote the object? Do they even do it at all? Some natlang examples would be fabulous. Thanks!

[Yng]

So, many ergative languages don't have a passive voice, choosing instead to have an antipassive. Now, what I'm wondering is how exactly do languages with an antipassive but no passive voice demote the agent or promote the object? Do they even do it at all? Some natlang examples would be fabulous. Thanks!

Why would they need to? Some ergative languages (Sumerian I think is one) can simply delete ergative arguments, resulting in just absolutive + verb. Word order can also be used to bring the absolutive to the beginning of the sentence. The absolutive is the object, and is the only required argument, so the ergative can be removed without a problem. The problem in these languages is when you want a word to have only an agentive, because as I understand it, in languages with an antipassive 'John-ABS eat' will be interpreted as 'John is eaten', 'John-ERG eat' as ungrammatical and 'John-ABS eat-ANTIPASSIVE', as 'John eats'.

I may've misunderstood, if so, tell me.

[Grimalkin]

Why would they need to? Some ergative languages (Sumerian I think is one) can simply delete ergative arguments, resulting in just absolutive + verb. Word order can also be used to bring the absolutive to the beginning of the sentence. The absolutive is the object, and is the only required argument, so the ergative can be removed without a problem.

Ok, that makes sense. Thanks.

I'd assume that most languages have a way of demoting or deleting the agent when it's unspecified. Though I suppose you could use stative verbs a lot of the time: e.g. the vase was broken could be a stative verb, but it's less clear cut when you have something like risks were taken. It's obvious that the speaker is deliberately leaving the agent out by using a passive construction. Can any natlangs not do that? (Sorry, slightly different question to the original one)

The problem in these languages is when you want a word to have only an agentive, because as I understand it, in languages with an antipassive 'John-ABS eat' will be interpreted as 'John is eaten', 'John-ERG eat' as ungrammatical and 'John-ABS eat-ANTIPASSIVE', as 'John eats'.

Yep that makes sense. That must be why antipassives are so rare in active-stative languages, because in those languages 'John-ERG slides' and 'John-ABS slides' would have different meanings (ok, the cases in active-stative would probably be called 'ergative' and 'accusative', not 'ergative' and 'absolutive) Which leads me on to another question... could an active-stative language have a passive voice? I'm wondering because my conlang is kinda active-stative, but doesn't have a valency-decreasing voice. If you want to omit the object, you use a dummy object or stick a detransivitiser on the verb. And I've yet to work out how I could demote or omit the agent. It's a bit tricky because case is only marked on pronouns so I can't just change the case and have done with it.

[Yng]

I'd assume that most languages have a way of demoting or deleting the agent when it's unspecified. Though I suppose you could use stative verbs a lot of the time: e.g. the vase was broken could be a stative verb, but it's less clear cut when you have something like risks were taken. It's obvious that the speaker is deliberately leaving the agent out by using a passive construction. Can any natlangs not do that? (Sorry, slightly different question to the original one)

I'm confused as to exactly what you're asking. I'd imagine most if not all natlangs have some way of getting rid of unwanted agents, as well as simply promoting patients, but not all of them do it by passive voice. If you had all of those act as stative verbs I'd imagine it'd be treated as a passive anyway... but if you had something like the example I gave before, which was:

John-ERG meat-ABS eat - John eats the meat
meat-ABS eat - meat is eaten
John-ERG eat-ANTIPASSIVE - John eats

Then that would accomplish the same thing without a passive.

Yep that makes sense. That must be why antipassives are so rare in active-stative languages, because in those languages 'John-ERG slides' and 'John-ABS slides' would have different meanings (ok, the cases in active-stative would probably be called 'ergative' and 'accusative', not 'ergative' and 'absolutive)

Weeeeell, I don't think that'd necessarily be the case. You could have John-ERG eat-ANTIPASSIVE, John-ABS eat-ANTIPASSIVE, couldn't you, and then John-ABS eat-ACTIVE which would be something like 'John is eaten'. I'm not familiar enough with any split-S natlangs to tell you if they do this, though.

Which leads me on to another question... could an active-stative language have a passive voice? I'm wondering because my conlang is kinda active-stative, but doesn't have a valency-decreasing voice. If you want to omit the object, you use a dummy object or stick a detransivitiser on the verb. And I've yet to work out how I could demote or omit the agent. It's a bit tricky because case is only marked on pronouns so I can't just change the case and have done with it.

A detransitiviser is a valency-decreasing voice, isn't it? If your system is something like this:

Transitive:
Subject: ERG. Object: ABS.

Intransitive:
Subject: ERG/ABS (ERG = more agent-like, ABS = more patient-like)

Then you can have a system like this:

'Antipassive' (detransitiviser):

Transitive > Intransitive.
Subject: ERG (Agent). (Optional Patient: ???)

'Passive':

Transitive > Intransitive.
Subject: ABS (Patient). (Optional Agent: ???)

Maybe? I'm not really sure. In my conlang, which is ergative with optional marking of 'subjects' as ERG or ABS depending on volition (in both transitive and intransitive contexts), verbs require only a patient argument, which is literally the patient (the thing the verb is done to), and so 'passive voice' can be accomplished simply by deletion of agentive arguments. A verb with only an agent is 'detransitivised' (I suppose, actually, it's kind of an antipassive voice).

[Bedelato]

So, many ergative languages don't have a passive voice, choosing instead to have an antipassive. Now, what I'm wondering is how exactly do languages with an antipassive but no passive voice demote the agent or promote the object? Do they even do it at all? Some natlang examples would be fabulous. Thanks!

Why would they need to? Some ergative languages (Sumerian I think is one) can simply delete ergative arguments, resulting in just absolutive + verb. Word order can also be used to bring the absolutive to the beginning of the sentence. The absolutive is the object, and is the only required argument, so the ergative can be removed without a problem. The problem in these languages is when you want a word to have only an agentive, because as I understand it, in languages with an antipassive 'John-ABS eat' will be interpreted as 'John is eaten', 'John-ERG eat' as ungrammatical and 'John-ABS eat-ANTIPASSIVE', as 'John eats'.

I may've misunderstood, if so, tell me.

I've always wondered what was so "anti" about the antipassive voice.

Here's my take on what you said:
In accusative langs, the subject (nominative) of an intransitive verb is an agent by default, and to make it a patient the passive voice is used.
But in ergative languages, apparently the subject of an intransitive verb (absolutive) will be interpreted by default as a patient, and the antipassive is used to make it into an agent.

Am I right or wrong here?

[Yng]

I've always wondered what was so "anti" about the antipassive voice.

Here's my take on what you said:
In accusative langs, the subject (nominative) of an intransitive verb is an agent by default, and to make it a patient the passive voice is used.
But in ergative languages, apparently the subject of an intransitive verb (absolutive) will be interpreted by default as a patient, and the antipassive is used to make it into an agent.

Am I right or wrong here?

In some ergative languages. This would only apply to ergative languages which equate absolutive with patient, like in English ergative verbs - The window broke -> He broke the window. I don't know if all ergative languages act like this or if some assign roles similarly to non-ergative languages but happen to mark nouns differently to nom-acc.

But I don't think this is necessarily how the antipassive works - as far as I know, its only requirement is that it produce a verb with a single argument which is in the absolutive. I vaguely remember a discussion once about how in some Australian languages, only absolutive arguments can be the syntactic pivot, or whatever, so in certain sentences the ergative has to be demoted for those reasons.

[Grimalkin]

Hmmm, this has given me food for thought and confirmed my suspicion that it's not gonna be so easy to make a good Erglang. Thanks for the replies, there'll probably be another barrage of questions coming your way tomorrow after I've had a good night's sleep

[Astraios]

Lakota is split-S, and it has a pseudo-passive (homophonous with the animate plural marker). If there is no agentive noun in the sentence, the verb can either be passive or have a third-person animate plural agent. It can then be treated as a stative verb, i.e. one without an overt agent, and can modify nouns.

John wóte. John ate.
John yútapi. John was eaten. / They ate John.

Mní kiŋ yatké. He drank the water.
Mní kiŋ yatkáŋpi. The water was drunk. / They drank the water.

Tȟaló kiŋ yukpáŋpi. The meat was ground. / They ground the meat.
Tȟaló yukpáŋpi kiŋ yúte. He ate the hamburger.


It could be argued that Lakota also has a pseudo-antipassive, in the form of the prefix wa-, which prevents the verb from taking an explicit patient, but gives it an implicit patient (i.e. "things/people") instead:

John yúte. John ate it.
John wóte. John ate things. / John had a meal. (wótA < wa-yútA )

Mní kiŋ yatké. He drank the water.
Wayátke. He drinks things. / He is an alcoholic.

Ókiye. He helped her.
Wawókiye. He helps people. / He is helpful. (the extra <w> is for euphony)

[Grimalkin]

I might just have to steal Lakota, actually. Olguma (my conlang) is Fluid-S so when there's only one core argument, the case that it's in (ergative or accusative) affects the meaning of the sentence. Although every verb with a volitional agent has a mandatory agent marker, so if there's only one core argument and no agentive marker, then it means that the argument has no volition -- there could be a way of fitting in a pseudo-passive contruction into this.

I think my 'detransivitizer' is essentially an antipassive, but it also pops up in 'participles' (which aren't really IE-style participles but they're verbal adjectives, so close enough) and some verbal nouns. This has inspired me to start posting bits of Olguma grammar in C&C now. It's still confusing though

[Vardelm]

I posted this link in the L&L Resources thread, but it's buried on pg 12. It may be useful here.

Typology of Ergativity by William McGregor


To understand ergative languages, I think it's really helpful to compartmentalize the topic into morphological, syntactic, and semantic ergativity. If you understand how ergativity works at those different levels, you can then look at glosses of different languages and see which type of ergativity that language has. It helps to compare a pure nom/acc lang to a pure erg/abs lang (which has all 3 types of ergativty, which is rare!), and then also to a morphologically ergative / syntactically accusative lang like Basque.

IMO, most of the confusion surrounding ergative langs stems from looking at partially ergative langs, not understanding what level of ergativity is displayed, and assuming that the language is a good example of "pure" ergativity when in fact it's not. It may be a good example of how ergative features may show up in languages, but that's different than understanding ergativity.

[Yng]

To understand ergative languages, I think it's really helpful to compartmentalize the topic into morphological, syntactic, and semantic ergativity. If you understand how ergativity works at those different levels, you can then look at glosses of different languages and see which type of ergativity that language has. It helps to compare a pure nom/acc lang to a pure erg/abs lang (which has all 3 types of ergativty, which is rare!), and then also to a morphologically ergative / syntactically accusative lang like Basque.

Yeah, that's what I was trying and failing to say before. Basque marks its nouns in a way that's ergative, but an intransitive sentence can still contain just a patient because it's syntactically accusative.

That's a very helpful link, though. Thanks.

[TomHChappell]

Now, what I'm wondering is how exactly do languages with an antipassive but no passive voice demote the agent or promote the object? Do they even do it at all? Some natlang examples would be fabulous.

See http://wals.info/feature/combined?id1=107&id2=108.
Here are 62 languages with no passive and no antipassive:
Abkhaz, Alamblak, Amele, Arapesh, Asmat, Awa Pit, Bagirmi, Barasano, Batak (Karo), Brahui, Burmese, Daga, Dani (Lower Grand Valley), Epena Pedee, Ewe, Guaraní, Igbo, Ika, Imonda, Ingush, Ju|'hoan, Kewa, Khasi, Kilivila, Kobon, Koromfe, Lakhota, Lezgian, Luvale, Makah, Maung, Maybrat, Meithei, Mixtec (Chalcatongo), Ndyuka, Ngiyambaa, Nivkh, Nunggubuyu, Oneida, Oromo (Harar), Otomí (Mezquital), Paamese, Pirahã, Rama, Rapanui, Samoan, Sango, Shipibo-Konibo, Suena, Supyire, Tagalog, Taiap, Tauya, Tiwi, Una, Usan, Wambaya, Wichita, Wichí, Yagua, Yimas, Yoruba.

According to http://wals.info/feature/combined?id1=100&id2=108, 2 languages in their database have ergative alignment of verbal person marking but have no antipassive constructions: Abkhaz and Macushi.
According to http://wals.info/feature/combined?id1=99&id2=108, 9 languages in their database have ergative-absolutive alignment of case marking of pronouns but have no antipassive constructions: Burushaski, Dani (Lower Grand Valley), Epena Pedee, Ingush, Kewa, Lezgian, Shipibo-Konibo, Tukang Besi, Una.
According to http://wals.info/feature/combined?id1=100&id2=107, 10 languages in their database have ergative alignment of verbal person marking and have no passive constructions: Abkhaz, Atayal, Canela-Krahô, Cavineña, Kapampangan, Konjo, Lak, Trumai, Uma, Yup'ik (Central).
According to http://wals.info/feature/combined?id1=98&id2=108, 14 languagses in their sample database have ergative-absolutive alignment of case marking of full noun phrases but have no antipassive constructions: Burushaski, Dani (Lower Grand Valley), Epena Pedee, Ika, Ingush, Kewa, Lezgian, Ngiyambaa, Paumarí, Shipibo-Konibo, Suena, Tukang Besi, Una, Wambaya.
According to http://wals.info/feature/combined?id1=99&id2=107, 15 languages in their sample database have ergative-absolutive alignment of case marking of pronouns and have no passive constructions: Bawm, Bribri, Chukchi, Dani (Lower Grand Valley), Epena Pedee, Gooniyandi, Ingush, Kewa, Ladakhi, Lezgian, Shipibo-Konibo, Trumai, Una, Wardaman, Zoque (Copainalá).
According to http://wals.info/feature/combined?id1=98&id2=107, 25 languages in their sample database have ergative-absolutive alignmentj of case marking of full noun phrases and have no passive constructions.



Only Chamorro and Trumai in their sample have both absolutive-ergative alignment of case marking of pronouns and ergative alignment of verbal person marking.
Only Lak, Trumai, and Yup'ik (Central) in their sample have both absolutive-ergative alignment of case marking of full noun phrases and ergative alignment of verbal person marking.
19 languages in their sample have both absolutive-ergative alignment of case marking of full noun phrases and absolutive-ergative alignment of case marking of pronouns.

Trumai is the only language in their sample that's ergative by all three features.
It has no passives, but they don't say whether or not it has antipassives.

Chamorro has both passives and antipassives.
Lak doesn't have passives but does have antipassives.
Yup'ik (Central) doesn't have passives but does have antipassives.


Abkhaz has neutral alignment of case marking of both full noun phrases and of pronouns, but has ergative alignment of verbal person marking.
Macushi has ergative alignment of verbal person marking. They don't say anything about its case-marking.

[Grimalkin]

Nice replies, thanks

Vardelm, I've downloaded the PDF and I'll read it when my brain will let me. I'm in the process of changing Olguma from a split-ergative language to a fully fledged fluid-S language. So obviously it will have ergativity, but not in the Basque or Georgian sense. Damn, I really need to read that PDF.

Tom, cheers for the stats. It's good to see there's so much variety in natlangs; it gives me plenty of wiggle room with Olguma, which I think has an antipassive but no passive voice. I'm leaning towards marking ergativity on verbs - when I've finished restructuring Olguma, most (maybe all) verbs will carry a kind of topic marker which agrees with the most agentive noun in the clause. At the moment, only volitional agents are marked on the verb. Expect a thread in C&C about Olguma's ergathingammywutsit system soon.

[Turama]

I'm confused as to exactly what you're asking. I'd imagine most if not all natlangs have some way of getting rid of unwanted agents, as well as simply promoting patients, but not all of them do it by passive voice. If you had all of those act as stative verbs I'd imagine it'd be treated as a passive anyway... but if you had something like the example I gave before, which was:

John-ERG meat-ABS eat - John eats the meat
meat-ABS eat - meat is eaten
John-ERG eat-ANTIPASSIVE - John eats

I do not understand why ergative languages can't have a passive voice. Because if it would be like in your example one could say "meat-AKK eat" in a nominative–accusative languages and it would mean "meat is eaten", but nevertheless most these languages have passives constructions and it mostly look like this "meat-NOM eat-PASSIVE". So I think, that theoretically ergative languages could have passive voice in transative verbs. Correct me if I understood something wrong.

Sorry if I formulate something wrong or if I missunderstood something in the post, I'm not so good in English.

[Morrígan]

I do not understand why ergative languages can't have a passive voice.

The problem is that, in from the perspective of Nominative-Accusative languages, Ergative languages are always using "passive voice".

In Nom-Acc alignment, the agent argument is the "Subject" (not a universally applicable idea), and in Nom-Acc passive constructions, the agent is omitted or demoted, and the patient argument becomes the subject.

In an idealized Erg-Abs language, the subject is the patient: in Chechen and other East Caucasian languages, the verb agrees in number and gender with the patient.

[Cúlro]

I do not understand why ergative languages can't have a passive voice.

The problem is that, in from the perspective of Nominative-Accusative languages, Ergative languages are always using "passive voice".

The same argument would mean that ergative verbs in English like 'to break' can't have passives: the window broke (intransitive) and the window was broken (passivised transitive) are different in meaning - only the passive voice specifically signals the existence of the agent. The intransitive form describes the change to the patient but is silent on the existence or not of the agent. Ergative languages can have a passive voice for this reason.

[Hallow XIII]

There are languages with ergativity and passive voices. There was a much newer thread on this you could have necro'd.

[Vardelm]

3+ years. Quite the necro!

The problem is that, in from the perspective of Nominative-Accusative languages, Ergative languages are always using "passive voice".

This is one way of putting it, and I think it might be a good way, but I'm not totally decided yet.

I'll rephrase it a bit. In a pure nom-acc language (that is, nom-acc in morphology, syntax, etc.), the default, unmarked voice is active. That is, the subject of both the intransitive and transitive tends to be more agent-like in nature. In order to form a sentence that is passive (make the subject be a patient or more patient-like argument), a more marked/derived sentence construction must be used.

A pure abs-erg language (erg abs in morphology, syntax, etc.) can be viewed as being the opposite. The default, unmarked voice is passive. The subject of both the intransitive and transitive tends to be more patient-like in nature. To form a sentence that is active (make the subject be an agent or more agent-like argument), a more marked/derived sentence construction must be used.

One has to realize that there really is no such thing as a 100% "pure" nom-acc or abs-erg language. You might be able to say that far more nom-acc languages come closer to "pure" nom-acc behavior than languages with abs-erg features do to "pure" abs-erg behavior. So, you end up with a majority of languages that are morphologically ergative showing nom-acc syntax, etc.

I do not understand why ergative languages can't have a passive voice.

2 issues here:

1) You have to define what you mean by "ergative language". Is this only morphological ergativity, or is it a combination of morphological & syntactic features? I think syntactic ergativity makes a large difference in how a language behaves regarding voice.

2) It's not that ergative languages, even ones with syntactic ergativity, CAN'T have a "passive" voice. It's just that languages that are morphologically ergative AND syntactically ergative are (probably) less likely to have a "passive" voice. If they do, it's probably performing a slightly different function than a passive voice in a syntactically accusative language.


Several points to illustrate this:

Intransitives in ergative languages often resemble a passive voice.
In ergative languages, the subject of ambitransitive verbs in intransitive sentences take the same thematic role as the patient or more patient-like argument of the transitive. Importantly, they may do so with no different in marking. I don't have real language examples at my finger tips, so take these arbitrary constructed examples based on Yng's post:

Johnza veburgen munchaka.
John-za ve-burgen munch-ak-a.
John-ERG DEF-meat.ABS eat-PERFECTIVE-3SingPatient
John ate the meat.

Veburgen munchaka.
Ve-burgen munch-ak-a.
DEF-meat.ABS eat-PERFECTIVE-3SingPatient
The meat eat was eaten.


Note how the only difference between the two sentences is the addition of the ergative agent, John. The actual translation of the intransitive might not be ideal, depending on the language. The passive in English often implies that there is an unspoken agent involved. Thus, it might be better to translate with something like a middle voice. Some verbs work well for this in English (as in "the bread is baking"), and some don't (such as "the meat ate").


Syntactically ergative languages can often drop the ergative agent w/ no additional marking
Again, the only difference between the examples above is the addition of the ergative agent in the transitive. That means the ergative argument can simply be dropped in the intransitive. A number of languages that are morphologically and syntactically ergative function this way. This means that they pattern in a manner that closely parallels the passive voice in English as in:

The meat was eaten by John.

The meat was eaten.


A "passive" voice in an ergative language might just be required to allow the ergative agent dropping
If the language in question does not allow verbs do simply drop a core argument without modification, you might get something like this:

Johnza veburgen munchaka.
John-za ve-burgen munch-ak-a.
John-ERG DEF-meat.ABS eat-PERFECTIVE-3SingPatient
John ate the meat.

Veburgen munchalaka.
Ve-burgen munch-al-ak-a.
DEF-meat.ABS eat-PASSIVE-PERFECTIVE-3SingPatient
The meat eat was eaten.


In this case, perhaps the language doesn't have ambitransitive verbs, or only very few. Verbs that can be transitive thus require an added marker (-al- in this example) in order to drop/demote the agent. This behaves differently than the English passive in that the patient isn't moving in the word order into a subject position necessarily. It also may or may not imply an unspoken agent. Even though it can function different, using the term "passive" might be the best term just because the agent is being dropped/demoted.

There may be better examples out there. Hopefully this does the job...

Again, though, a number of (many? a majority? all?) languages that are both morphologically and syntactically ergative don't bother with requiring added morphosyntactic elements in order to drop the ergative agent so they function like the 1st example. Their standard, unmarked transitive & intransitive sentences pivot around the patient, so a "passive" voice as it is understood in English and similar nom-acc languages just isn't absolutely needed.

Because if it would be like in your example one could say "meat-AKK eat" in a nominative–accusative languages and it would mean "meat is eaten", but nevertheless most these languages have passives constructions and it mostly look like this "meat-NOM eat-PASSIVE".

Many languages have a required noun case for the subject of a language. In English, you are not going to say "me freezing", for example. If you do say that, there's some humor being expressed through an obviously ungrammatical construction. So, languages like that might require some morphosyntactic operation in order to accomodate a non-agent-like noun in the subject slot.

So I think, that theoretically ergative languages could have passive voice in transative verbs. Correct me if I understood something wrong.

I think the main thing you need to look at is how the intransitive subject is interpreted/translated in an ergative language compared to the transitive. That alone will show why ergative languages tend to not have a passive voice, or at least perhaps not as much as nom-acc languages and serving slightly different functions.

[Cúlro]

To be fair, the necromancer is a non-native speaker and it was first (and so far only) post.

Welcome, Arhama!

[Vardelm]

The same argument would mean that ergative verbs in English like 'to break' can't have passives: the window broke (intransitive) and the window was broken (passivised transitive) are different in meaning - only the passive voice specifically signals the existence of the agent. The intransitive form describes the change to the patient but is silent on the existence or not of the agent. Ergative languages can have a passive voice for this reason.

This is a very good example of the point I was trying to make:

The actual translation of the intransitive might not be ideal, depending on the language. The passive in English often implies that there is an unspoken agent involved. Thus, it might be better to translate with something like a middle voice. Some verbs work well for this in English (as in "the bread is baking"), and some don't (such as "the meat ate").

...

In this case, perhaps the language doesn't have ambitransitive verbs, or only very few. Verbs that can be transitive thus require an added marker (-al- in this example) in order to drop/demote the agent. This behaves differently than the English passive in that the patient isn't moving in the word order into a subject position necessarily. It also may or may not imply an unspoken agent. Even though it can function different, using the term "passive" might be the best term just because the agent is being dropped/demoted.

[Terra]

Arhama, read this thread : viewtopic.php?f=7&t=41677