Ergativity for Novices

[bradrn]

The only suggestion I'd have so far is to add a section about "lexical ergativity", although that's really just a discussion of accusative, ergative, unaccusative, and unergative verbs.

I’d love to add a section on that! But I don’t know anything about the area — do you have any articles you can link to on this subject?

Just the McGregor article.

Now I’m confused — as far as I can see, the article only has half a page on lexical ergativity (in section 2.2), much of which is taken up by examples. (You are talking about the article called Typology of ergativity, right?) But from what you’re saying, it sounds like the article has much more information about lexical ergativity than I’m seeing. Am I looking in the wrong place in the article, or is that small section really all there is?

[Vardelm]

Now I’m confused — as far as I can see, the article only has half a page on lexical ergativity (in section 2.2), much of which is taken up by examples. (You are talking about the article called Typology of ergativity, right?) But from what you’re saying, it sounds like the article has much more information about lexical ergativity than I’m seeing. Am I looking in the wrong place in the article, or is that small section really all there is?

That small section on "lexical-semantic" ergativity is all there is, unfortunately. I doubt there is much other literature describing this type of "ergativity" since, as McGregor mentions "some linguists to the use of the term ergativity in relation to lexical alternations like these".

In my view, a language would qualify as "semantically ergative" when a majority - or perhaps just a large number relative to other languages - of the verbs are ergative and/or unaccusative. That would mean that, for most verbs, subjects of intransitives tend to be more patientive, corresponding to direct objects of transitives. That's pretty much the definition of ergativity, which is why McGregor (I think) takes the view that using the term "ergative" is appropriate here.

[bradrn]

Now I’m confused — as far as I can see, the article only has half a page on lexical ergativity (in section 2.2), much of which is taken up by examples. (You are talking about the article called Typology of ergativity, right?) But from what you’re saying, it sounds like the article has much more information about lexical ergativity than I’m seeing. Am I looking in the wrong place in the article, or is that small section really all there is?

That small section on "lexical-semantic" ergativity is all there is, unfortunately. I doubt there is much other literature describing this type of "ergativity" since, as McGregor mentions "some linguists to the use of the term ergativity in relation to lexical alternations like these".

Pity… I would have loved to know more about this subject!

In my view, a language would qualify as "semantically ergative" when a majority - or perhaps just a large number relative to other languages - of the verbs are ergative and/or unaccusative. That would mean that, for most verbs, subjects of intransitives tend to be more patientive, corresponding to direct objects of transitives. That's pretty much the definition of ergativity, which is why McGregor (I think) takes the view that using the term "ergative" is appropriate here.

‘Semantic ergativity’ sounds like a very interesting concept, and one that I’d like to explore in much greater depth — again, it’s a pity there aren’t any materials on this subject!

[Richard W]

Does the concept of ergativity have any connection with languages where voice and tense are frequently unmarked, such as Thai, Lao and related SVO languages? For example, /kai^B2 kin^A2/ can have a whole range of meanings including 'The chicken (/kai^B2/) is eating (/kin^A2/)' and 'The chicken has been eaten'. /kin/ is transitive prototypically. (I've used the 4-row Gedney box to mark tones.)

[chris_notts]

Does the concept of ergativity have any connection with languages where voice and tense are frequently unmarked, such as Thai, Lao and related SVO languages? For example, /kai^B2 kin^A2/ can have a whole range of meanings including 'The chicken (/kai^B2/) is eating (/kin^A2/)' and 'The chicken has been eaten'. /kin/ is transitive prototypically. (I've used the 4-row Gedney box to mark tones.)

This is just ambi-transitivity. Plenty of otherwise accusative languages have many S=P ambitransitives (intransitive subject = transitive object), including English. English even has some examples where either the transitive subject or object can be the intransitive dsbject, much like those SE Asian languages you mention:

I cooked the chicken
The chicken is cooking / cooked
I cooked

I think it may also be a bit misleading to characterise those languages as SVO, because they are heavily topic prominent so the first argument is a T not an S. Fronting of topics, unmarked for role and regardless of role, is common and may even be more heavily grammaticalised than the subject. See, e.g. https://en.m.wikipedia.org/wiki/Topic-p ... t_language

[bradrn]

Does the concept of ergativity have any connection with languages where voice and tense are frequently unmarked, such as Thai, Lao and related SVO languages? For example, /kai^B2 kin^A2/ can have a whole range of meanings including 'The chicken (/kai^B2/) is eating (/kin^A2/)' and 'The chicken has been eaten'. /kin/ is transitive prototypically. (I've used the 4-row Gedney box to mark tones.)

I agree with chris_notts: generally, ergativity has nothing to do with voice, tense or transitivity. If you read my previous chapter, I actually did address this:

(The mention of ambitransitive verbs is worth elaborating on. Some ambitransitive verbs have S=A, for instance I go there / I go; others have S=O, for instance I break it / It breaks. Despite the fact that the former verb appears ‘accusative’, whereas the later appears ‘ergative’, the number and type of ambitransitive verbs has no relation to whether a language is mainly accusative or ergative. For instance, Latin is accusative, whereas Dyirbal is ergative, but both have a nearly complete lack ambitransitive verbs; Fijian is not ergative, but almost all verbs in Fijian are ambitransitive, and there is a nearly equal split between S=A and S=O ambitransitive verbs, with only a slight bias (3%) towards S=A.)

Still, I wouldn’t be surprised if accusative languages with ambitransitive verbs are biased towards S=A, and ergative languages with ambitransitive verbs are biased towards S=O.

[Vardelm]

Still, I wouldn’t be surprised if accusative languages with ambitransitive verbs are biased towards S=A, and ergative languages with ambitransitive verbs are biased towards S=O.

I would agree & would file this under "semantic/lexical ergativity". I've done a little searching today & haven't found anything yet about the distribution of ergative verbs, unaccusative verbs, etc. among languages with ergativity. (Hey, linguistics majors, is that maybe an interesting topic for a paper?) A ZBBer once mentioned that morphologically ergative languages tend to have many/most intransitive verbs where the subject tends to be a theme/patient role. I think this was in a discussion about Basque (chris_notts, was that you maybe?). I haven't seen that confirmed in any articles, though.

[chris_notts]

Still, I wouldn’t be surprised if accusative languages with ambitransitive verbs are biased towards S=A, and ergative languages with ambitransitive verbs are biased towards S=O.

I would agree & would file this under "semantic/lexical ergativity". I've done a little searching today & haven't found anything yet about the distribution of ergative verbs, unaccusative verbs, etc. among languages with ergativity. (Hey, linguistics majors, is that maybe an interesting topic for a paper?) A ZBBer once mentioned that morphologically ergative languages tend to have many/most intransitive verbs where the subject tends to be a theme/patient role. I think this was in a discussion about Basque (chris_notts, was that you maybe?). I haven't seen that confirmed in any articles, though.

This is certainly true for some. In Basque, the situation is a bit complicated. Transitivity is overtly marked most of the time by the choice of auxiliary, and it's certainly true that for verbs which can take either there seem to be a lot of S=P pairs, although I have no statistics. Some complications that spring to mind, although I haven't spoken much Basque in a long time:

1. There are a lot noun + light verb combinations which basically always take the transitive auxiliaries, so as far as I'm aware such an alternation is not possible. E.g. hitz egin = word make = speak.

2. On the other hand, some progressive constructions take an intransitive aux even when combined with a transitive verb, and I'm not sure what this does to the ambitransitives. I'm thinking specifically of the ari construction, which also shows divergence case marking.

[chris_notts]

Another thing worth mentioning about Basque is that the "infinitive" form of the verb, the default form, is identical to what would be called the past / passive participle in other languages. So that verb form, when followed by the intransitive auxiliary izan (to be), is structurally quite similar to a passive or a perfect in English anyway:

hil dut = die 3sg-have-1sg = I killed it
hil da = die 3sg-be = it died
hila da = die-def 3sg-be = it/he/she is dead

The infinitive is ambiguous between active and passive functions. In adnominal and adjectival use it always has a passive-like function, but when used as a complement it is often active. Compare:

jan nahi dut = eat want/wish 3sg-have-1st = I want to eat, NOT I want to be eaten

[jal]

Great start, thanks! One typo:

serv-us
slave-NOM
veni-t
come-3s
The master comes


JAL

[bradrn]

Great start, thanks! One typo:

serv-us
slave-NOM
veni-t
come-3s
The master comes

I’d better fix that — thanks for finding this jal!

[bradrn]

Morphological ergativity: Case-marking, agreement and word order

What is morphological ergativity?

The term morphological ergativity refers to those areas of ergativity which are directly reflected in the morphology of a language: that is, ergative case-marking and ergative verbal agreement. (Dixon occasionally terms this area of ergativity intra-clausal ergativity, which is certainly a more precise term, but which is also slightly too unwieldy for my taste.) Ergative case-marking refers to the situation in which one noun case is used for A arguments and another is used for S=O arguments, while ergative verbal agreement refers to the situation in which one set of verbal agreement affixes is used for A arguments and another is used for S=O arguments.

A very similar — though rare — phenomenon is ergative word order, in which one position in the sentence is used for A arguments and another is used for S=O arguments. It can be hard to know how to categorise this: you might think that, as a matter of syntax, this should be treated as a facet of ‘syntactic ergativity’ rather than ‘morphological ergativity’, but it doesn’t seem particularly similar to the other phenomena treated as syntactically ergative (though further investigation is needed). On the other hand, like the other phenomena covered in this post, word order is concerned about identifying which NP corresponds to S, A and O; for this reason, I will cover it here.

Ergative case-marking

Case-marking in ergative languages works in a pretty similar manner to case-marking in non-ergative languages, except for the fact that the main two cases are ergative (for A) and absolutive (for S and O) rather than (say) nominative and accusative. They have much the same morphological possibilities too — cases may be expressed as affixes, as in the Dyirbal examples mentioned in the previous section (‘What is ergativity?’):

ŋuma-∅

father-ABS

banaga-nʸu

return-NONFUT

Father returned

ŋuma-∅

father-ABS

yabu-ŋgu

mother-ERG

bura-n

see-NONFUT

Mother saw father

yabu-∅

mother-ABS

banaga-nʸu

return-NONFUT

Mother returned

Or they may be expressed with particles or adpositions, as with Tongan (example from Dixon, glosses not given):

naʻe lea [ʻa Tolu]_S Tolu spoke
naʻe lea [ʻa e talavou]_S The young man spoke
naʻe tāmateʻi [ʻa e talavou]_O [ʻe Tolu]_S Tolu killed the youth

(Note that a Tongan sentence begins with the verb, here either intransitive lea ‘speak’ or transitive tāmateʻi ‘kill’, followed by the arguments of that verb, introduced using either ʻe ‘ergative marker’ or ʻa ‘absolutive marker’.)

As with non-ergative languages, it is much less common for languages to use particles or adpositions rather than affixes for case-marking, but the examples above demonstrate that both are possible.

Markedness in case-marking

An important topic in ergative case-marking is that of markedness. However, before discussing this topic, I want to note that this term has two closely related meanings, which should be disambiguated. The first is that of formal markedness, which relates to which of a system of affixes (or particles etc.) has the simplest or most basic form; often, the formally unmarked affix is zero, whereas all the other affixes in the system have a non-zero realisation. The second is that of functional markedness, which relates to which of a system of affixes etc. has the ‘more usual, more normal’ meaning (Dahl 1985). For example, one could make a case that the English 3s agreement affix -s is formally marked (all other agreement affix(es) are zero), but functionally unmarked (most normal sentences use -s).

With regards to markedness in case-marking, one generalisation is very clear: If any case is formally and functionally unmarked, it will be the case which covers S. That is, the unmarked case will be nominative (in a nominative-accusative language) or absolutive (in an ergative-absolutive language). For instance, in the Dyirbal examples above, the ergative case has a nonzero suffix whereas the absolutive case is zero. This generalisation holds true for all known ergative languages, and almost all nominative-accusative languages (the exceptions will be covered later).

The major consequence of this fact is that many phenomena appear ‘reversed’ compared to a nominative-accusative language like English, since the ‘main’ unmarked transitive argument is O rather than A. For instance, accusative languages often have a passive voice which promotes the O argument to the ‘more important’ A argument; by contrast, ergative languages often have an ‘antipassive’ voice which promotes the A argument to the ‘more important’ O argument. Another example: in a construction like I saw it and ∅ laughed, in English the omitted argument is assumed to be A, but some ergative languages assume that argument is O instead. Such phenomena will be dealt with in the section on syntactic ergativity later. In my opinion, this ‘reversal’ is one of the most frequently confusing things about ergative languages, so it is important to be aware of this at an early stage.

Another important consequence of markedness is related to the function of the more marked case: in both nominative and ergative languages, the more marked case — that is, accusative or ergative — is more likely to have functions beyond simply marking the A or O status of its NP. For instance, the Latin accusative case also covers ‘place to which there is motion’, ‘extent in time and space’ and certain ‘adverbial relations’ (Kennedy 1962, quoted in Dixon); similarly, the ergative also often covers locative, instrumental, genitive or other meanings (although there are some languages, such as Basque, which do have a ‘purely’ ergative case).

Finally, it seems that if the verb agrees with only one case role, it will agree with the unmarked case. For instance, in nominative-accusative languages, the verb often agrees with the nominative argument only, but there are no languages in which the verb agrees with the accusative argument only. Similarly, in ergative-absolutive languages, the verb may agree with the absolutive argument only, but never with the ergative argument only (although Päri may be an exception; see below for details).

(Aside: There do exist so-called ‘marked nominative’ systems, which are clearly nominative-accusative but have the accusative unmarked rather than the nominative. These systems are pretty much restricted to some northern African languages from the Afroasiatiatic and Nilotic families, although they are also found in the Yuman family of California. These systems are in some ways midway between ergative and ‘normal’ nominative-accusative languages: they could alternately be viewed as ergative languages where the marked ergative case has been extended to cover S role — Dixon notes that ‘this is not an easy terminological question’. These systems, although nominative-accusative in form, show several phenomena more characteristic of ergative languages: for instance, the citation form of words is often the unmarked accusative, and so is the complement of the word ‘to be’. There are some other languages in which phonological change has led to an apparent marked nominative system — for instance, in some older Germanic languages, the PIE accusative *-am has disappeared, whereas the nominative *-os has not — but in these systems, the nominative is still functionally unmarked. Another ambiguous example is English, in which the accusative pronoun me is functionally unmarked relative to the nominative pronoun I — yet English is never termed a marked nominative. It appears that a language must have an accusative which is consistently and unambiguously both formally and functionally unmarked to qualify as a ‘marked nominative’)

Ergative verbal agreement

I am sure that we are all familiar with verbal agreement (also sometimes called ‘cross-referencing’): the phenomenon where the verb or verbal auxiliary contains affixes which provide information about the person, number, gender or other attribute of one or more arguments. Often, these agreement markers provide evidence of morphological accusativity or ergativity.

First, an example from Swahili:

m-li-anguka You all fell down
tu-li-anguka We fell down
m-li-tu-ona You all saw us
tu-li-wa-ona We saw you all

Ignoring the past tense prefix li-, it is clear that Swahili has one set of agreement markers for S and A, and another for O:

	S/A	O
2p	m-	-wa-
3p	tu-	-tu-

Additionally, S and A always appear before the tense marker, whereas O appears afterwards. Thus it is clear that Swahili verbal agreement shows an ‘accusative’ pattern.

Similarly, there are also languages with an ‘ergative’ system of verbal agreement. An example is Sacapultec Maya:

š-at-ak-ek You entered
š-∅-ak-ek He/she entered
š-∅-a꞉-čʼiy-aŋ You hit him/her
š-at-ri-čʼiy-aŋ He/she hit you

Ignoring the completive aspect marker š-, and the transitivity markers -ek and -aŋ, this is clearly an ergative system rather than an accusative one, with one set of agreement markers for S and O, and another for A:

	S/O	A
2s	-at-	-a꞉-
3s	-∅-	-ri-

As with Swahili, the agreement markers have a particular order: S and O always appear first, and A appears afterwards. Additionally, note that 3s is unmarked, but only with S/O: this coincides with the earlier discussion about markedness, where it was mentioned that it is always the absolutive that is unmarked relative to the ergative. Thus it is clear that Sacapultec verbal agreement shows an ‘ergative’ pattern.

There are also some languages in which the verb only agrees with one of its arguments. Some familiar examples are the nominative-accusative languages of Indo-European, in which the verb agrees with S and A but not O. Similarly, there are some ergative languages, such as Avar and Canelo-Krahô (from the Jê family), in which the verb agrees with S and O but not A. This was briefly discussed earlier along with marking: if the verb only agrees with one argument, it will be that argument which corresponds to the less marked case (nominative or absolutive).

Another common phenomenon is for languages to have a synchronically unanalysable set of agreement markers for all A+O combinations, and another set of prefixes for S. In some ways, this could be considered as a type of tripartite marking. But in all known languages (or at least the vast majority) which do this, the A+O prefixes developed from recognisably separate A and O affixes, one of which was the same as (or nearly identical to) S.

Dixon writes that he knows of no language which has a tripartite system of agreement across all tenses and aspects. However, he states that the Mayan language Chorti comes close: in the imperfective aspect, there is a separate set of agreement markers for each of S, A, O, but in the perfective aspect, the O markers are used for S as well.

It is worth noting that ergativity in verbal agreement seems to be rarer than ergative case-marking. McGregor points out that according to WALS, 17% of a sample of 190 languages have ergative case-marking (https://wals.info/feature/98A#2/25.5/148.9), whereas only 5% of a sample of 380 languages have ergative verbal agreement (https://wals.info/feature/100A#2/17.0/274.0, although I have a suspicion that this dataset under-represents areas which contain particularly many languages with ergative agreement).

The relationship between ergative verbal agreement and case-marking

For the most part, languages — whether accusative, ergative or neither — tend to use either verbal agreement (head-marking) or case-marking (dependent-marking), but not both (double-marking). For instance, Sacapultec uses only verbal agreement, whereas Dyirbal uses only case-marking. However, the use of both is common enough that their co-occurrence should be discussed.

The most surprising thing about double-marking languages is that not all combinations are equally plausible. For instance, accusative agreement and ergative case-marking is relatively common, whereas ergative agreement and accusative case-marking is practically non-existent. The various cases can be summarised in a table (based on Ellen Woolford’s paper Ergative Agreement Systems and Amy Deal’s ergativity handbook):

	nom-acc agreement	agr with both erg and abs	agr with abs only	agr with erg only
unequivocal nom-acc case	English	unattested	(Canela)	unattested
morphologically unmarked case	Swahili	Jacaltec	unattested	(Halkomelem)
unequivocal ergative case	Warlpiri	Kabardian	Hindi	(Päri)

The various cases here can be summarised as a universal: if a language has accusative case-marking, then it cannot have any form of ergative verbal agreement. However, there are a few subtleties to be aware of here:

• The Jê language Canela seems to be a partial exception to this universal. Canela has nominative-accusative case-marking on pronouns, and no case-marking on other nouns (the same system as English). However, the verb agrees with S and O — an ergative agreement system in a (partially) accusative case-marking system.
• There seem to be no languages in which nouns have no case-marking but verbs agree with only the absolutive argument. However, I do have a suspicion that this may be an accidental gap: as far as I am aware, there are very few languages in general where verbal agreement is only with the absolutive argument, and it could be an accident that all such languages have case-marking.
• For the reasons described above under ‘Markedness’, there should be no languages where the verb agrees with the ergative argument alone. (Such a situation would be akin to an accusative language in which agreement is only with the accusative argument.) There do seem to be at least two partial exceptions to this general rule, but both are unusual cases: in Halkomelem, the verb only agrees with a third person ergative argument, while in Päri (described in more detail below) the verbal agreement is only present in a particular topicalisation construction.

I would finally like to note that there are in fact good theoretical reasons for these patterns. One of them — that of markedness — has already been used to explain the gap in the final column. However, a fuller discussion of the theory behind the rest of the table — and why the universal above is true — will have to wait until split ergativity and the animacy hierarchy are discussed later; the discussion above is probably sufficient for now.

Two separate cross-referencing systems

(This section isn’t about ergativity per se, aside from the fact that the agreement systems here all equate S and O in some way, but is merely a fascinating area which Dixon covers and which I wanted to include.)

In the majority of languages with verbal agreement, the verb agrees with at least two arguments; it is actually families such as Indo-European which are less usual, in having verbal agreement with only one argument. (According to WALS, there are 193 languages in which the verb agrees in person with both A and O, but only 97 in which the verb agrees in person with only one argument.) However, there also exist more complex agreement systems, in which two completely separate agreement systems are used at the same time:

• In Gavião, a clause contains one or more verbs, as well as an auxiliary. There may be an explicit NP in S or A function; if there isn’t, the auxiliary takes a prefix indicating the person or number of the S or A argument. Similarly, if the verb is transitive, there may be an explicit NP in O function; if there isn’t, the verb takes a prefix indicating the person and number of the O argument. If the verb is not transitive, it takes a prefix agreeing with S instead. Thus a Gavião clause contains two agreement markers: one, on the auxiliary, agrees with S or A, while the other, on the verb, agrees with S or O. Interestingly, both sets of agreement prefixes have almost exactly the same form, despite the fact that the mark separate sets of arguments.
• In Jarawara, there are two types of transitive construction: the ‘A-construction’, used ‘where the A argument continues an established topic’, and the ‘O-construction’, used ‘where the O argument continues an established topic’ (Dixon). There are also several suffixes attached to the auxiliary or verb; many of these suffixes agree in gender with one argument of the verb. In the A-construction, these suffixes agree with the A argument, whereas in the O-construction, these suffixes agree with the O argument. The verb also contains another agreement system, in the form of two ‘obligatory bound pronomial slots’ in a nominative-accusative pattern: the first slot agrees with the O argument, and the second slot agrees with the S or A argument.
• The Papuan language Koiari also has two agreement mechanisms: according to Dixon, ‘verbs in Koiari have suffixes indicating number … of S or O, followed by a tense/aspect suffix which also encodes some information concerning the person or number of S or A’.

These systems are quite different from more ‘normal’ systems of agreement, such as those of Swahili and Sacapultec outlined above. The verbal agreement affixes in those languages work as a single system: even though there are two slots, they both encode the same information about the arguments, the only difference between the slots being that one agrees with A and the other agrees with O. By contrast, the languages described in the above list each have two completely separate agreement systems: each set of agreement affixes may convey completely different sorts of information (as in Jarawara or Koriari), or mark the same argument twice (present in all three examples).

Ergative word order

So far, we have seen that ergative marking of arguments can manifest in case-marking, through a case system with ergative and absolutive cases, as well as in verbal agreement, through the use of separate agreement affixes for ergative and absolutive arguments. So it seems only natural to ask: can ergativity manifest in word order as well? After all, there are many languages which use word order rather than case or agreement to mark which argument is which.

Firstly, what would it mean for word order to show an ergative pattern? In languages such as English, the word order in intransitive clauses is subject then verb, and the word order in transitive clauses is agent then verb then patient. Dixon represents such a system as SV/AVO, and I will use this same notation here. From this description, we could say that English has accusative word order: S and A are in the same position in the sentence, and O is at a different place. An ergative word order, then, would be a system like VS/AVO, where S and O are in the same position, and A is different.

Note that A and O must be on different sides of the verb for a language to have ergative word order (although see the example of Sanuma below); for instance, if a language has SV/AOV word order, then it cannot really be said to have either accusative or ergative word order. This is why many otherwise fully ergative languages show no evidence of ergative word order: ergative languages are usually verb-initial or verb-final (Manning 1994), and of course such languages cannot be said to have either ergative or accusative word order.

Another logically possible case is tripartite word order, where all three arguments are treated differently. This is only possible if S is on one side of the verb, and A and O are on the other, e.g. SV/VAO. As far as I am aware, there is no language with such a system.

As it turns out, we know of no language which marks ergativity solely through word order (with the partial exception of the Austronesian language Tolai, which will be covered later). However, there are some languages where ergativity is shown by other means, but also have clearly ergative word order:

• The Nilotic language Päri (mentioned briefly above) has unmarked order SV in intransitive clauses, and OVA in transitive clauses. A is marked with ergative case, whereas O is unmarked (or possibly marked with a null absolutive marker). There is additionally a construction which is used to topicalise the A argument: in this construction, A is fronted (changing the order to AOV), the ergative case-marker is removed, and an affix agreeing with the agent is added. For instance:
  
  

  ùbúr

  Ubur

  á-túuk

  COMPL-play

  
  Ubur played
  
  

  jòobì

  buffalo

  á-kèel

  COMPL-shoot

  ùbúrr-ì

  Ubur-ERG

  
  Ubur shot the buffalo
  
  

  ùbúr

  Ubur

  jòobì

  buffalo

  á-kèel-é

  COMPL-shoot-3sA

  
  Ubur shot the buffalo
  
  This means that Päri has intransitive word order SV and transitive word order OV, and A can go either before or after the OV unit. It is thus clear that Päri has ergative word order.
• The Carib language Kuikúro has a similar pattern to Päri: the basic word order is SV for intransitive clauses and either OVA or AOV for transitive clauses.
• The Nadëb language of Brazil has the basic order SV/OAV; an alternate word order is VS/AVO. Here, the basic ‘unit’ of word order is V/AV, with S and O going either before or after this unit; thus S and O are treated the same with respect to word order, and A is treated differently.
• The Mayan language Huastec has an ergative system of verbal agreement, as is common in Mayan languages (e.g. Sacapultec, as described above). The most frequent word order in Huastec is the ergative system VS/AVO, although Dixon notes that ‘variation is possible’.
• The Yanomami language Sanuma has basic word order SV/AOV. As mentioned above, this word order is neither accusative nor ergative. However, when peripheral NPs — denoted X — are taken into account, the word order becomes XSV/AXOV, which is clearly ergative.

Finally, a warning from Dixon about determining ergativity purely from word ordering:

It will be seen that all the languages quoted here as having some kind of ergative pattern in their constituent ordering show other ergative features,
at the level of intra-clausal marking. Indeed, since constituent order fulfils a wide variety of pragmatic as well as grammatical functions, we should hesitate to characterise a language as ‘ergative’ on the basis of constituent order alone.

[Pabappa]

So does this mean that there are languages with bipersonal marking on verbs, where the agent appears closer to the root of the verb than the patient does? Somewhere I thought I'd read that the object appears closer in all known languages.

[bradrn]

So does this mean that there are languages with bipersonal marking on verbs, where the agent appears closer to the root of the verb than the patient does? Somewhere I thought I'd read that the object appears closer in all known languages.

This is actually a point which I would like to know much more about, but I haven’t found any information about this anywhere. But, I think the answer to your question is yes: in languages like Sacapultec (and also Abkhaz I believe, although I don’t have access to Dixon’s book just at the moment), the subject/patient are further from the verb, whereas the agent is closer.

[jal]

The next section (on split ergativity) looks like it will be a bit longer than the previous ones, so it could take a bit more time to write.

Since it's a two-headed beast, you could split it (no pun intended :)) between fluid-S and split-S?


JAL

[bradrn]

The next section (on split ergativity) looks like it will be a bit longer than the previous ones, so it could take a bit more time to write.

Since it's a two-headed beast, you could split it (no pun intended ) between fluid-S and split-S?

That’s only one small part of split ergativity. I also need to cover splits according to the animacy hierarchy (which in and of itself could prove lengthy), splits on TAM, splits on main vs subordinate clauses, combinations of splits, and one or two interesting case studies which I want to include. Active-stative systems could well prove to be the shortest part of the whole thing.

[jal]

That’s only one small part of split ergativity. I also need to cover splits according to the animacy hierarchy (which in and of itself could prove lengthy), splits on TAM, splits on main vs subordinate clauses, combinations of splits, and one or two interesting case studies which I want to include. Active-stative systems could well prove to be the shortest part of the whole thing.

One introductory post, and then one post per subject as per the above? Anyway, great series so far, I'm rooting for more :).


JAL

[bradrn]

That’s only one small part of split ergativity. I also need to cover splits according to the animacy hierarchy (which in and of itself could prove lengthy), splits on TAM, splits on main vs subordinate clauses, combinations of splits, and one or two interesting case studies which I want to include. Active-stative systems could well prove to be the shortest part of the whole thing.

One introductory post, and then one post per subject as per the above?

That could certainly work, but then I’d be worried about some sections being too small… Anyway, I’ve already written the introduction, so I’m just thinking I’ll just split it in two at some point between subsections if I feel it’s getting too long. I think that would work — would you (as a reader) be fine with that?

Anyway, great series so far, I'm rooting for more .

Why, thank you! I’m glad you like it.

[jal]

would you (as a reader) be fine with that?

But of course! Anything is better than nothing :).


JAL

[Xwtek]

Also, it's worth pointing out the ergativity in English. (i.e. My reading vs Troy's destruction by the Greeks). In English, action nominals seems to have a split intransitive alignment. Compare above with "My singging the Marseillaise"