Word order in an ergative language

[jal]

I was wondering about the typical word order in an ergative language. Is the ABS argument in a transitive sentence considered the subject? So in a an SVO ergative language, the word order is patient-verb-agent? And what about languages like Hindi that have ergativity only in the past tense?


JAL

[Vijay]

No, the ERG argument in a transitive sentence would be considered the subject. In an intransitive sentence, the ABS argument would be considered the subject, though. So AFAIK, word order works pretty much the same way in ERG-ABS languages as it does in NOM-ACC languages.

[jal]

No, the ERG argument in a transitive sentence would be considered the subject. In an intransitive sentence, the ABS argument would be considered the subject, though. So AFAIK, word order works pretty much the same way in ERG-ABS languages as it does in NOM-ACC languages.

Ok, so the agent is still considered the subject, thanks. I also just saw a video on Basque that confirms that at least in Basque the ergative argument comes first.


JAL

[Vardelm]

No, the ERG argument in a transitive sentence would be considered the subject. In an intransitive sentence, the ABS argument would be considered the subject, though. So AFAIK, word order works pretty much the same way in ERG-ABS languages as it does in NOM-ACC languages.

Sort of, but not quite.

It depends on whether the language is syntactically accusative or ergative. The example you gave is for a syntactically accusative language, where the ergative argument (agent) is in the same position as the intransitive subject. In a syntactically ergative language, the absolutive argument (patient) of a transitive sentence will be in the same location as the intransitive subject. This of course assumes that the language uses word order to help determine argument role.

IMO, statements that word order works the same in ergative languages as accusative languages, without qualification, leads to an enormous amount of confusion regarding ergativity. Most languages that have ergative features, such as (especially) morphologically ergative case marking, have accusative syntax. However, not all of them do, and you can't call that syntax "ergative" without creating confusion. The best course is to describe the morphology, syntax, and semantics of a language as accusative or syntax separately.

I always, always, always link to the following PDF in discussions about ergativity. Read it and understand it to understand ergativity.

https://www.researchgate.net/profile/Wi ... f78852.pdf


If that link doesn't work for you, simply Google "typology ergativity mcgregor".

[jal]

I always, always, always link to the following PDF in discussions about ergativity. Read it and understand it to understand ergativity.

The link worked fine, thanks! Starting reading in 1-2-3...


JAL

[Vardelm]

The link worked fine, thanks! Starting reading in 1-2-3...

Cool. Let me know if you have questions. I've been referring to that article for years now, and I've found it to be incredibly informative and applicable.

[Cedh]

I second Vardelm's reading suggestion. A very good paper that should be required reading for every conlanger who plans to dabble with ergativity.

Note also that the vast majority of ergative languages are either verb-initial or verb-final. Verb-medial basic word orders such as SVO are extremely rare in combination with any kind of ergativity, and even thought to be impossible by some linguists. (Can't remember the source for that claim though.)

[cromulant]

Note also that the vast majority of ergative languages are either verb-initial or verb-final. Verb-medial basic word orders such as SVO are extremely rare in combination with any kind of ergativity, and even thought to be impossible by some linguists.

Paumarí may be an exception.

https://en.wikipedia.org/wiki/Paumar%C3 ... age#Syntax
http://tinyurl.com/jc9oplt

[Cedh]

Yes, that seems to be a counterexample, and a very interesting variant of split ergativity too. Thanks a lot for the link!

[Cedh]

Double post with a summary of the Paumarí data for those who haven't followed cromulant's link, or can't see the relevant pages on Google Books, because it's really interesting:

Paumarí basically has the following three transitive constructions:

• subject-ERG verb object
• subject object-ACC verb
• object-ACC verb subject

These constructions are selected on a pragmatic basis, with the ergative SVO construction being the default, but available only if both subject and object are already established topics. New agents are typically introduced in intransitive clauses before they appear in an agentive role, and new patients are typically introduced using one of the two accusative constructions (SOV or OVS). This is striking (a) because it's pragmatics-based split ergativity, which I haven't come across before, and (b) because the language is ergative exactly and only in those circumstances (SVO word order) where ergativity is least expected, typologically. The system still feels very coherent though, because it's always only the immediately pre-verbal argument which is morphologically marked.

It's tempting to suggest an alternative interpretation here where both the ergative and accusative markers are not case suffixes on the noun, but rather voice prefixes on the verb. But AFAICS there are two arguments against such an analysis: First, both subjects and objects may be dropped, which results in the case marker being dropped too but the verb retaining transitive morphology. Second, pronouns coreferential with the object (which are required with plural animate objects) exhibit case agreement, so that the accusative marker appears twice: subject object-ACC 3PL.A-ACC verb. Also, it's possible that other words, e.g. adverbs, may intervene between the pre-verbal noun phrase and the verb, although I have found no examples of this in the paper.

[Vardelm]

Thanks for the links, crom! I looked at them, and that's a really nice summation, Cedh. It's definitely a VERY interesting syntax! It's also a good example of syntactic ergativity, which is rare, in that it has default VS / SVO word order.

[jal]

Cool. Let me know if you have questions. I've been referring to that article for years now, and I've found it to be incredibly informative and applicable.

Excellent article! Thanks for sharing. Funny to see that speculation about past/present split, and origins of ergativity are in line with what I myself had come up with (even though the current conlang I'm working on has an ergative that's clearly derived from a passive) :).

Note also that the vast majority of ergative languages are either verb-initial or verb-final. Verb-medial basic word orders such as SVO are extremely rare in combination with any kind of ergativity, and even thought to be impossible by some linguists.

Unfortunately, the paper didn't touch this, so the question arises: why would this be? If looking at the proposed origins of ergativity (in the paper), nothing seems to indicate a necessity for ergative constructions to be verb initial or verb final. Of course, this could be pure chance, but in that case I wouldn't think that SVO would be "thought to be impossible by some linguists". On what basis do they think so?


JAL

[Vardelm]

Note also that the vast majority of ergative languages are either verb-initial or verb-final. Verb-medial basic word orders such as SVO are extremely rare in combination with any kind of ergativity, and even thought to be impossible by some linguists.

Unfortunately, the paper didn't touch this, so the question arises: why would this be? If looking at the proposed origins of ergativity (in the paper), nothing seems to indicate a necessity for ergative constructions to be verb initial or verb final. Of course, this could be pure chance, but in that case I wouldn't think that SVO would be "thought to be impossible by some linguists". On what basis do they think so?

I think this is bunk, since there's the example of Paumari:

Paumarí basically has the following three transitive constructions:

• subject-ERG verb object
• subject object-ACC verb
• object-ACC verb subject

These constructions are selected on a pragmatic basis, with the ergative SVO construction being the default, but available only if both subject and object are already established topics.

Granted Paumari has the second two orders, but those have restrictions. WALS lists Paumari as primarily SVO, and nothing is listed as ergative/OVS for noun phrase case marking. WALS says Parï has OVS with verbal person marking that is ergative, but nothing for SVO. Since ergative languages are uncommon anyway, I tend to think that this is more a case of sample size than being "impossible". The development path of some ergativity (especially nominalization?) may lead away from verb-medial structures, such as:

thing.HEADNOUN otherthing.GENITIVE action.VERB > SOV


Other paths to ergativity could, I think, still lead to veb-medial order, such as passive order in English leading to OVS.

[Travis B.]

Unfortunately, the paper didn't touch this, so the question arises: why would this be? If looking at the proposed origins of ergativity (in the paper), nothing seems to indicate a necessity for ergative constructions to be verb initial or verb final. Of course, this could be pure chance, but in that case I wouldn't think that SVO would be "thought to be impossible by some linguists". On what basis do they think so?

I personally tend to take the view that any view that languages must behave a certain way because such-and-such author's favorite theory says they must behave that way is probably a load of bunk.

[Cedh]

Like Vardelm said, the main reason for the rarity of ergative SVO languages seems to be that the common development paths for ergativity lead away from SVO order:

The alleged dispreference for ergative alignment in SVO languages is attributed to the typical positioning of the oblique constituents which are considered to constitute the source of ergative nominal marking. In SVO languages oblique constituents are typically placed after the object (SVOX), i.e. on the opposite side of the verb than the transitive A or intransitive S. In SOV and V1 languages, on the other hand, oblique constituents are generally located on the same side of the verb as the A or S, typically in one of the following X-positions: SOXV, SXOV, VSOX, VSXO, VOSX, VOXS. The most commonly postulated source of ergative nominal marking is a former passive agent.
[...]
Another source of ergative nominal marking [...] is that of oblique instrumental NPs in transitive clauses with covert As, in structures such as NP-instr NP-acc V.
[...]
Assuming that reanalyses are the product of ambiguity in surface data, ambiguity between a passive and an active interpretation is much more likely to arise in the case of passive constructions when the passive agent occurs on the same side of the verb as the transitive A than those in which the passive agent is on the opposite side of the verb than the A . And analogously in the case of an instrumental and transitive A interpretation of instrumental NPs in transitive clauses. Accordingly, both the passive agent and instrumental source of the ergative marker are seen to be more compatible with basic V3, V1 or free word order than with SVO or OVS.

In the survey performed for the paper cited above, Siewierska actually finds that 8 out of 77 verb-medial languages in her sample have ergative marking on nouns (10%), but only 3 out of 77 languages (4%) have ergative marking on pronouns.¹ For 6 out of 77 verb-medial languages (8%), ergative alignment qualifies as the "dominant alignment type".²

¹) If only languages with non-neutral alignment in nouns or pronouns respectively are taken into account, the figures are 33% (8 out of 24) for nouns and 9% (3 out of 33) for pronouns. These are the numbers she gives in her paper.
²) 'Dominant alignment' is defined as "a) the alignment of the majority of parts of speech; b) the sole non-neutral alignment; c) the alignment of nouns rather than pronouns; d) in cases of triple splits, the left-most alignment on the following hierarchy: hierarchical > active > tripartite > ergative > accusative".

[jal]

Thanks y'all, that makes sense.


JAL

[cromulant]

In the survey performed for the paper cited above, Siewierska actually finds that 8 out of 77 verb-medial languages in her sample have ergative marking on nouns (10%), but only 3 out of 77 languages (4%) have ergative marking on pronouns.¹ For 6 out of 77 verb-medial languages (8%), ergative alignment qualifies as the "dominant alignment type".²

Then ergative marking on verb medial languages...really isn't that rare at all? Where did this idea come from that it was impossible? Am I missing something?

[Travis B.]

Some people confuse less common with impossible, especially when predictions made by theories come into play.

[Curlyjimsam]

Siewierska's paper seems to have been largely ignored as far as I can tell - loads of people have repeated the "ergative languages are never SVO" claim since it was published, though in some cases admitting one or two counterexamples.

It doesn't help that she doesn't actually say, as far as I can see, which languages are ERG+SVO - which makes her claim rather harder to evaluate. We can't necessarily assume, in a 200-odd language sample, that the author hasn't simply misidentified a language's alignment and/or basic word order in some instances. I get the impression she may have undercounted the number of languages with active case systems, for instance (a not uncommon error - the literature often describes examples of such languages as "ergative").