Re: Vardelm's Scratchpad (NP: Devani OVS order)

[Vardelm]

Funny, I could have sworn I already wrote and sent a reply to the post before this… should I be worried?

It just gets worse with time. Wait until you're really old.... like 30.

I’m not sure I understand. Does this mean you’re sticking with the below-quoted system?

Note that the above current version just adds/changes a voice suffix to the predicate verb and swaps the position of the active and stative arguments. The new version below adds the same suffix. However, the referents change positions, but the inverse argument marking stays in the same order as the direct, unlike the original system.

Yes, I'll be sticking with that system (99% sure now?). I like how the argument marking is playing out with that. What I'm really changing with this last post was the rules for the MVCs. When I mentioned this:

After some more thinking, I don't like the structure where you could use a verb with active direct voice and then a stative indirect voice and the arguments would simply be reversed for one of them. It's complicated, ugly, & doesn't seem natural.

I was specifically referring to this bit:

ak'ijuqawi lazhimbu qam magawa ja tumbuwi ak'iju-qa-wi break-ACT.DIR-SGT lazhi-mbu sitting-STA.INV qam that.which.3P.ACT magawa is.unknown ja that.which.3P.STA tumbu-wi is.chair-the.SGT Someone broke the chair and was sitting on the chair. Someone broke the chair by sitting on it.

You can see that the active and stative assignments agree between those examples, although they are in a different order. I'm not wild about the order difference, but that's OK.

Notice that lazhimbu is in the stative inverse voice. That means its argument order should be stative ja - active qam if it were appearing in a sentence all by itself. However, because the sentence starts with the active ak'ijuqawi, the arguments are ordered as active qam - stative ja. It's that inconsistency that I found "complicated".

If so, I can say that this one seems natural enough to me, and not complicated or ugly at all. Certainly, it’s natural enough to have been used in a bunch of Austronesian languages. (Mostly Indonesian ones, since as you noted, Philippine-type languages have a rather more elaborate system.)

Very good!

It helps if you think of the determiners as marking ‘direct’ and ‘oblique’ rather than ‘active’ and ‘stative’ arguments.

This comment is interesting to me. I'm not sure this would work. For active verbs, the "active" determiners would be direct and the "stative" determiners oblique. However, I think that would be reversed for the stative verbs. The "stative" determiners would be direct and "active" determiners oblique. What I have is - if I analyze correctly - 2 nominative-accusative system that are mirror images of each other in terms of which "case" they use to mark what.

Also, I find myself somewhat suspicious of the term ‘multi-verb predicate’. The whole point of calling things ‘multi-verb constructions’, as opposed to a more specific term, is that we don’t want to care about whether a given phrase is a single ‘predicate’ or not, or for that matter a single ‘clause’ or not. I do wonder if there’s any difference at all between a ‘multi-verb predicate’ and a ‘serial verb construction’, or if they’re not just two names for the same thing.

Fair enough. I think using "MVC" works just as well. I favor using a term like this due to the difficulty of pinning down what constitutes a "serial verb" or doesn't.

I also note that the relevant verbs are contiguous in all your examples, so perhaps that should be an additional requirement, especially considering how common non-contiguous SVCs are.

That's a REALLY good point. One constant during this development has been that I have referred to a string of multiple, contiguous verbs at the front of a clause/sentence as being SVCs, MVCs, or whatever term is preferred. I'll make sure to note that in a future grammar.

I see no problem with instrumental meanings — the usual construction is something along the lines of <subject> take <instrument> <verb>, which would be fine in Jin. You wouldn’t be able to add an instrumental argument to a transitive verb, but it would work well enough for intransitives.

Good point. I was thinking mostly of transitives.

One more thing to consider: are there any semantic differences between your MVCs/SVCs and the corresponding coordinate constructions? Sometimes there aren’t with MVCs (e.g. in English). But there do tend to be some, particularly when they’re a common construction, and especially when there’s a word-order change as there is in Jin. (Not that I have a reference for that, or even a particularly solid reason, but it at least makes intuitive sense.)

I'm not sure yet, but I would lean towards "no". If I can figure out a general pattern, rule, etc. that would describe the differences, I'd be more likely to adopt it. Once thing I don't want to get into with Jin (or my other current conlangs) is detailing every last specific usage in the language. I think it's awesome when conlang authors go into that much detail, but I definitely don't want to here. These are intended to be proto-languages, and there are other parts of world-building I would like to eventually focus on.

[bradrn]

When I mentioned this:

After some more thinking, I don't like the structure where you could use a verb with active direct voice and then a stative indirect voice and the arguments would simply be reversed for one of them. It's complicated, ugly, & doesn't seem natural.

I was specifically referring to this bit:

ak'ijuqawi lazhimbu qam magawa ja tumbuwi ak'iju-qa-wi break-ACT.DIR-SGT lazhi-mbu sitting-STA.INV qam that.which.3P.ACT magawa is.unknown ja that.which.3P.STA tumbu-wi is.chair-the.SGT Someone broke the chair and was sitting on the chair. Someone broke the chair by sitting on it.

You can see that the active and stative assignments agree between those examples, although they are in a different order. I'm not wild about the order difference, but that's OK.

Notice that lazhimbu is in the stative inverse voice. That means its argument order should be stative ja - active qam if it were appearing in a sentence all by itself. However, because the sentence starts with the active ak'ijuqawi, the arguments are ordered as active qam - stative ja. It's that inconsistency that I found "complicated".

I’m not entirely sure I understand… so how would you translate that sentence now?

It helps if you think of the determiners as marking ‘direct’ and ‘oblique’ rather than ‘active’ and ‘stative’ arguments.

This comment is interesting to me. I'm not sure this would work. For active verbs, the "active" determiners would be direct and the "stative" determiners oblique. However, I think that would be reversed for the stative verbs. The "stative" determiners would be direct and "active" determiners oblique.

Yes, you are correct here; sorry for misanalysing it.

What I have is - if I analyze correctly - 2 nominative-accusative system that are mirror images of each other in terms of which "case" they use to mark what.

On the other hand, this doesn’t work either, because if I understand correctly, both the ‘active’ and the ‘stative’ determiners can be used for the intransitive argument. So really this system is neither accusative nor ergative nor entirely Austronesian, but something entirely new. (Though I maintain that Austronesian alignment is closest to yours.)

One more thing to consider: are there any semantic differences between your MVCs/SVCs and the corresponding coordinate constructions? Sometimes there aren’t with MVCs (e.g. in English). But there do tend to be some, particularly when they’re a common construction, and especially when there’s a word-order change as there is in Jin. (Not that I have a reference for that, or even a particularly solid reason, but it at least makes intuitive sense.)

I'm not sure yet, but I would lean towards "no". If I can figure out a general pattern, rule, etc. that would describe the differences, I'd be more likely to adopt it. Once thing I don't want to get into with Jin (or my other current conlangs) is detailing every last specific usage in the language. I think it's awesome when conlang authors go into that much detail, but I definitely don't want to here. These are intended to be proto-languages, and there are other parts of world-building I would like to eventually focus on.

…OK, that’s fair. It makes me sad, but it’s fair.

On the other hand: have you considered the diachronic possibilities of well-developed SVCs? A key characteristic of SVCs is that they have a really strong tendency to both lexicalisation and grammaticalisation. By ‘lexicalisation’, I mean that more symmetric SVCs tend to turn into idiomatic compounds; by ‘grammaticalisation’, I mean that more asymmetric SVCs tend to turn into semantically bleached clitics or even affixes. And, as usual, these developments are easiest to figure out when you have a good understanding of how words combine into SVCs, and what those SVCs mean. I would say that it’s worth thinking about the semantics of specific cases at least, if only for this purpose. Besides, I wasn’t asking for anything really detailed; something like ‘here are five constructions where a word takes on a specialised meaning’ is fine. (That’s basically as much depth as I’ve gone to in my languages.)

Also, I’ll note another difference between lexicalisation and grammaticalisation. As you mention, the former usually consists of very specific combinations of words with idiomatic and more or less unpredictable meanings. And, as you mention, this may well be a bit too much effort to go to for a mere protoconlang. However, grammaticalisation is itself a process of semantic bleaching and generalisation, so is far more amenable to description in a reference grammar format. Any good resource on SVCs (including all the ones I’ve linked) should have plenty of examples of such asymmetric SVCs, where one word is mostly fixed but has a highly general meaning.

(The discussion above is very much SVC-specific, insofar as that’s a sensible category to distinguish: MVCs as a more general category include less tightly linked constructions, which I presume would not be under nearly as much pressure to evolve. But as I said, your constructions are in fact fairly tightly linked, not being analysable as paratactic coordination in the same way as some other MVCs are.)

[Vardelm]

I’m not entirely sure I understand… so how would you translate that sentence now?

That will probably be answered in the next post I'm planning. For now, that Jin sentence is just ungrammatical (according to the updated rules I posted) so it would just be formed a little different.

Yes, you are correct here; sorry for misanalysing it.

On the other hand, this doesn’t work either, because if I understand correctly, both the ‘active’ and the ‘stative’ determiners can be used for the intransitive argument. So really this system is neither accusative nor ergative nor entirely Austronesian, but something entirely new. (Though I maintain that Austronesian alignment is closest to yours.)

No problem at all, sir! Here's a bit more detail on what I'm seeing:

Active verb determiners:
Active verb intransitive subject 3P = qam
Active verb agent 3P = qam
Active verb patient 3P = ja

Stative verb determiners:
Stative verb intransitive subject 3P = ja
Stative verb agent 3P = ja
Stative verb patient 3P = qam

So that looks like 2 accusative systems to me, again with the different just being the mirrored marking.

On the other hand: have you considered the diachronic possibilities of well-developed SVCs? ... Besides, I wasn’t asking for anything really detailed; something like ‘here are five constructions where a word takes on a specialised meaning’ is fine. (That’s basically as much depth as I’ve gone to in my languages.)

All of that sounds pretty doable, and actually within the scope of what I'd like to do with these since they are proto-langs! Let's see if I can get these 4 proto-langs to the point I'd like them to be before moving to descendent languages!

[bradrn]

I’m not entirely sure I understand… so how would you translate that sentence now?

That will probably be answered in the next post I'm planning. For now, that Jin sentence is just ungrammatical (according to the updated rules I posted) so it would just be formed a little different.

I’m thinking, could this perhaps work?

lazhi-nza

sitting-STA.DIR

ak'iju-un-i

break-STA.INT-SGT

qam

that.which.3P.ACT

magawa

is.unknown

ja

that.which.3P.STA

tumbu-wi

is.chair-the.SGT

That way, ‘sit’ gets the person as its agent and the chair as its patient, while ‘break’ gets the chair as its patient. This is how most languages with SVCs do it.

(This, by the way, is why dependent-marking languages don’t have SVCs quite as commonly as head-marking ones do. When it does happen, they tend to be a bit lax around things like making the case-markers all line up. The NE Ambae grammar [open-access again!] has a good overview of this sort of thing.)

Yes, you are correct here; sorry for misanalysing it.

On the other hand, this doesn’t work either, because if I understand correctly, both the ‘active’ and the ‘stative’ determiners can be used for the intransitive argument. So really this system is neither accusative nor ergative nor entirely Austronesian, but something entirely new. (Though I maintain that Austronesian alignment is closest to yours.)

No problem at all, sir! Here's a bit more detail on what I'm seeing:

Active verb determiners:
Active verb intransitive subject 3P = qam
Active verb agent 3P = qam
Active verb patient 3P = ja

Stative verb determiners:
Stative verb intransitive subject 3P = ja
Stative verb agent 3P = ja
Stative verb patient 3P = qam

So that looks like 2 accusative systems to me, again with the different just being the mirrored marking.

Yes, this does indeed look accusative… though not very accusative, since it feels somewhat artificial to delineate two qams and two jas like this. I’ll have to keep on thinking.

[Vardelm]

I’m thinking, could this perhaps work?

lazhi-nza

sitting-STA.DIR

ak'iju-un-i

break-STA.INT-SGT

qam

that.which.3P.ACT

magawa

is.unknown

ja

that.which.3P.STA

tumbu-wi

is.chair-the.SGT

That way, ‘sit’ gets the person as its agent and the chair as its patient, while ‘break’ gets the chair as its patient. This is how most languages with SVCs do it.

For the way that I have Jin structured right now, no, that doesn't work. Yes, it could if the notions of agency & subject-hood were more closely linked like they are in a typical accusative language. I'll try to run through my (broken?) logic from the ground up.

lazhi

sitting.STA.INT

ja

that.which.3P.STA

ndalu-wi

is.man-the.SGT

The man is sitting.

Lazhi is - by default - stative intransitive voice, so I marked it as sitting.STA.INT since there is no voice suffix to change the voice from the default. Since this is the intransitive voice, that means the subject - ndaluwi "the man" - is in a state of sitting. Because it's a steady, unchanging, non-dynamic state, the determiner ja is used. With stative verbs (in Jin), the steady state is more associated with subject-hood.

lazhi-nza

sitting-STA.DIR

ja

that.which.3P.STA

ndalu-wi

is.man-the.SGT

qam

that.which.3P.ACT

tumbu-wi

is.chair-the.SGT

The man is sitting on/in the chair.

Lazhi now gets the stative direct suffix, making it transitive. The man is still in the steady state of sitting; he's the stative subject. The only change here is that the state is relative to something: the chair.

The second argument for a verb with stative direct voice is the stimulus or location of the state. (Perhaps you could also add source or origin to the list of thematic relations, which seem a bit fuzzy in cases like this.) In this example, it's the chair. I look at it as the chair (or whatever) is almost the cause of the state. At the least, the chair enables the state. As such, it might be ever so slightly higher on a sort of agency/dynamicity hierarchy, allowing the active determiner qam to be used. If nothing else, since state-hood is closely linked with subject-hood for stative voices, the active pronouns & determiners just become the "other" marking.

ak'iju

broken.STA.INT

ja

that.which.3P.STA

tumbu-wi

is.chair-the.SGT

The chair is broken.

Meanwhile, let's start with a simple state; the chair is broken. Again, ak'iju is default stative intransitive voice, so it takes a stative ja argument. However, notice that in the transitive example above, the chair has the active qam argument, so we can't add this construction to that one. Here's what we can do:

ak'iju-j-i

broke-ACT.INT-SGT

qam

that.which.3P.ACT

tumbu-wi

is.chair-the.SGT

The chair broke.

We add the active intransitive voice to ak'iju, which makes it a dynamic event. The chair changes its state from unbroken to broken. It becomes broken. Dynamicity is the difference between Jin's active and stative verbs & voices. Since it is an active voice, it uses an active qam argument.


Now we can construct the form I would use for your example.

lazhi-nza

sitting-STA.DIR

ak'iju-j-i

broke-ACT.INT-SGT

ja

that.which.3P.STA

ndalu-wi

is.man-the.SGT

qam

that.which.3P.ACT

tumbu-wi

is.chair-the.SGT

The man was sitting on/in the chair and it broke.

This satisfies the rules I posted, in particular the one about not using the other dynamicity (active vs stative) until after a transitive had been used to define the arguments. In this example, lazhinza is active, so you can't use a stative verb until a transitive active verb is used. However, lazhinza is already transitive due to having the active direct voice suffix, and so we can go ahead and add ak'ijuji to the predicate, and it uses the active qam argument as its subject.

The implication here is - of course - that the man was the cause of the chair breaking. We could strengthen that implication by using another active voice that is intransitive: the active passive.

lazhi-nza

sitting-STA.DIR

ak'iju-zhis-i

broke-ACT.PAS-SGT

ja

that.which.3P.STA

ndalu-wi

is.man-the.SGT

qam

that.which.3P.ACT

tumbu-wi

is.chair-the.SGT

The man was sitting on/in the chair and it was/became broken (by him).

The passive voices are used when there is an agent that is left unstated. Compare with the intransitive voices where there is no - or at least much less - implication of an agent.

What is not allowed by my current rules is using a transitive voice of both active and stative dynamicity. There can be 2 or more active transitive verbs, or 2 or more stative transitive verbs, but not a mix.

There are a few more constructions I have in mind that are similar to the above, and that will be the focus of that next post I mentioned (and still have to write).

Yes, this does indeed look accusative… though not very accusative, since it feels somewhat artificial to delineate two qams and two jas like this. I’ll have to keep on thinking.

Maybe think in these terms:

ja = Closer to a more steady, unchanging state.

qam = Closer to a dynamic, changing state or an agent that is causing that change.

If a ja argument is changing state, it's because there is an agent that is causing that change and is considered even more dynamic. If ja is used as an agent, it's because the verb is focusing on the subject's change of state and the agency of the ja argument is considered less important. (To my knowledge so far, this only happens in the active indirect voice.)

[Vardelm]

Jin: Adverbs & Results


Adverbs

As I described in this post and this one, the linker al- can be used to form adjunct phrases. I also showed adverbs, but the suffix -(n)un has been dropped.

zhama-hi eat-SGT qam that.which.3P.ACT ndalu-wi man-the.SGT ja that.which.3P.STA capitu-ma fruit-the.PLT al-miba of.with.LNK-quick The man ate the fruit and (he was) quick. The man ate the fruit with quickness. The man ate the fruit quickly.

The way I view the linker right now is that when attached to a verb link miba "quick", the subject of that verb is assumed to be the subject of the main clause. In this case, that's qam ndaluwi "the man". The linked verb's subject is gapped. Miba is a stative intransitive verb, so ordinarily it would require a stative ja subject. However, because its subject is gapped, the active/stative requirement is dropped and it refers to qam ndaluwi as mentioned.

zhama-hi eat-SGT qam that.which.3P.ACT ndalu-wi man-the.SGT ja that.which.3P.STA capitu-ma fruit-the.PLT al-miba-lah of.with.LNK-quick-STA.IMP The man ate the fruit quickly.

There is also an option to use the linker with a verb in the impersonal voice. Here, miba appears in the stative impersonal voice. Using the impersonal describes the sentence as a whole rather than the sentence's subject. Al-mibalah means the event as a whole didn't last much time rather than the man acted with any particular haste. The difference is slight and the 2 constructions could probably be used interchangeably.

miba-lah quick-STA.IMP zhama-hi eat-SGT qam that.which.3P.ACT ndalu-wi man-the.SGT ja that.which.3P.STA capitu-ma fruit-the.PLT Quickly, the man ate the fruit.

Impersonal verbs can appear in the predicate as part of a MVC. When they are the first verb in the series, they also describe the sentence's event as a whole. Since impersonal verbs don't take arguments, they can be used without regards to which of the verbs are active or stative or what voices are used.

uñgayub-i strangle-SGT miba-lah quick-STA.IMP mimut-aq-i dead-ACT.DIR-SGT qam that.which.3P.ACT ñgonda-hi is.warrior-the.SGT ja that.which.3P.STA umbabi-wi is.merchant-the.SGT The warrior quickly strangled and killed the merchant.

uñgayub-i strangle-SGT mimut-aq-i dead-ACT.DIR-SGT miba-lah quick-STA.IMP qam that.which.3P.ACT ñgonda-hi is.warrior-the.SGT ja that.which.3P.STA umbabi-wi is.merchant-the.SGT The warrior strangled and quickly killed the merchant.

When an impersonal verb follows another verb in an MVC, it describes that verb.



Results

uñgayub-i strangle-SGT mimut dead qam that.which.3P.ACT ñgonda-hi is.warrior-the.SGT ja that.which.3P.STA umbabi-wi is.merchant-the.SGT The warrior strangled the merchant and he (the merchant) is dead.

A stative intransitive verb can be used to show a result. Because a stative intransitive requires a stative argument, we know that mimut "dead" is referring to the merchant.

uñgayub-i strangle-SGT mimut-alah dead-STA.IMP qam that.which.3P.ACT ñgonda-hi is.warrior-the.SGT ja that.which.3P.STA umbabi-wi is.merchant-the.SGT The warrior strangled the merchant to death.

An impersonal voice can also be used to indicate a result.




More Breaking Chairs

Last post we had this example:

Now we can construct the form I would use for your example.

lazhi-nza

sitting-STA.DIR

ak'iju-j-i

broke-ACT.INT-SGT

ja

that.which.3P.STA

ndalu-wi

is.man-the.SGT

qam

that.which.3P.ACT

tumbu-wi

is.chair-the.SGT

The man was sitting on/in the chair and it broke.

To that, we can add a few more constructions.

lazhi-q-i sitting-ACT.DIR-SGT ak'iju-q-i broke-ACT.DIR-SGT qam that.which.3P.ACT ndalu-wi is.man-the.SGT ja that.which.3P.STA tumbu-wi is.chair-the.SGT The man sat down on and broke the chair.

The act of sitting here has become active, so it's now a dynamic change of state: the man became seated/sat down on the chair. When he did so, he also broke the chair.

ak'iju-q-i broke-ACT.DIR-SGT lazhi-mbis-i sitting-STA.PAS-SGT qam that.which.3P.ACT ndalu-wi is.man-the.SGT ja that.which.3P.STA tumbu-wi is.chair-the.SGT The man broke the chair by it being sat-in.

The first verb is now "broke", which is active. Because of that, lazhi "sit" is put into the stative passive "to be sat upon". This allows it to still use the stative ja argument and doesn't violate the rule against an active transitive and stative transitive verb in the same MVC. Because the passive is used, it implicates another actor, which of course is the man.

ak'iju-q-i broke-ACT.DIR-SGT lazhi-lah sitting-STA.IMP qam that.which.3P.ACT ndalu-wi is.man-the.SGT ja that.which.3P.STA tumbu-wi is.chair-the.SGT The man "sittingly" broke the chair. The man broke the chair by sitting.

Here, the impersonal is used. Since it follows ak'ijuqi "break", it describes that verb and becomes a method of how the verb was carried out.



This brings us to an example from several posts ago:

jau⁵

have

jan⁴

person

co⁵-laan⁶-zo²

sit-broken-PERV

zoung¹

CL

dang³

chair

Someone has broken the chair by sitting on it (Cantonese)

I see a few ways to form this sentence.

ak'iju-q-i broke-ACT.DIR-SGT lazhi-lah sitting-STA.IMP qam that.which.3P.ACT magawa is.unkown ja that.which.3P.STA tumbu-wi is.chair-the.SGT Someone broke the chair sitting.

ak'iju-q-i broke-ACT.DIR-SGT qam that.which.3P.ACT magawa is.unkown ja that.which.3P.STA tumbu-wi is.chair-the.SGT al-lazhi-lah of.with.LNK-sitting-STA.IMP Someone broke the chair by sitting.

ak'iju-q-i broke-ACT.DIR-SGT qam that.which.3P.ACT magawa is.unkown ja that.which.3P.STA tumbu-wi is.chair-the.SGT al-lazhi-nza of.with.LNK-sitting-STA.DIR qasim 3P.ACT Someone broke the chair by sitting on it.

[Vardelm]

Devani: OVS order = ergativity violation?

I mentioned that I was thinking about making Devani's comment portion of a sentence OVS. Now I'm seriously considering it.

mechani the mida paramam mechani scholar the TOP mida he.3P.ANIM paramam read Regarding the scholar: he is reading.

samhain the ne malkusuvam tujari samhani ceremony the TOP ne 2P.INF m-alkus-uv-am 3P-anoint.IRS-TRS.VOL-3P.VOL.VIS tujari priest For the ceremony, the priest will anoint you.

The examples show intransitive word order is SV, while transitive would be OVS. Based on just word order, this would make Devani have ergative syntax. This doesn't take into account pivots, etc. which could be added ergative features. I haven't thought about those yet.

I like this, but..... Note that the -am suffix (for 3rd person, volitional, visual evidence) agrees with the intransitive subject (theh scholar) and the transitive agent (the priest). This is accusative agreement. So I think this might be an issue. My understanding is that syntactically ergative languages have ergative marking. Devani has no case marking at all, so ergative case marking isn't a factor. Languages also can have ergative case marking but accusative verbal agreement, so those aren't linked. My question then is do any syntactically ergative languages have accusative verbal agreement?

[bradrn]

The examples show intransitive word order is SV, while transitive would be OVS. Based on just word order, this would make Devani have ergative syntax. This doesn't take into account pivots, etc. which could be added ergative features. I haven't thought about those yet.

I like this, but..... Note that the -am suffix (for 3rd person, volitional, visual evidence) agrees with the intransitive subject (theh scholar) and the transitive agent (the priest). This is accusative agreement. So I think this might be an issue. My understanding is that syntactically ergative languages have ergative marking. Devani has no case marking at all, so ergative case marking isn't a factor. Languages also can have ergative case marking but accusative verbal agreement, so those aren't linked. My question then is do any syntactically ergative languages have accusative verbal agreement?

This is an interesting case, because generally speaking, people don’t seem to classify ergative word order as ‘syntactic ergativity’ — it’s considered merely a way of individuating the arguments, on a par with case-marking and verbal cross-reference. Nonetheless, all languages I’m aware of with ergative word order also have some other form of ergative marking, whether it be case-marking or cross-referencing (or both).

However, there is some flexibility. The simplest options would be to either move the intransitive subject after the verb, or add an ergative case-marker. But also, note that most of the time, your topic will be an argument of the verb: sentences with structures S TOP V, A TOP O V or O TOP V A will all be common. If you have an ergative case-marker, this can let you reduce the amount you use it. Päri, for instance, uses the ergative case-marker only when A is not topicalised; my own conlang Hlʉ̂ is very similar, with the ergative obligatory only when O is topicalised. Luwo, closely related to Päri, has a more complex system (Storch 2014)

Clause type	Case-marking	Function
SV, OVA	only A is case-marked	unmarked
AO[V=A]	no case-marking but A is cross-referenced on the verb	unmarked
VS/VAO	both S and A are case-marked	sequential actions and events
AVO	none	object focus
OAV	none	negation

[Vardelm]

Nonetheless, all languages I’m aware of with ergative word order also have some other form of ergative marking, whether it be case-marking or cross-referencing (or both).

Yeah, that's what I was thinking it would be.

The simplest options would be to either move the intransitive subject after the verb, or add an ergative case-marker.

No case markers is a design goal for this language, so that's out. Looks like I'll be choosing between SV/SVO and VS/OVS. Jin is VS while Dwarven & Yokai are SV. According to WALS, the overwhelming majority or languages have SV order, so it seems like SV/SVO may be the way to go. Time to stare blankly at a ceiling for a while until I settle on either route.

[bradrn]

The simplest options would be to either move the intransitive subject after the verb, or add an ergative case-marker.

No case markers is a design goal for this language, so that's out.

It’s worth noting that the ergative is a very minimal sort of case-marker: it can be as small as a single phoneme, and is used only on transitive subjects — and not even on all transitive subjects, or even most of them. For reference, I can find only 30 instances total of the ergative case-marker in my copy of Storch’s Luwo grammar, which is ~300 pages long, and furthermore most of those instances are in the chapter on ergativity.

Looks like I'll be choosing between SV/SVO and VS/OVS. Jin is VS while Dwarven & Yokai are SV. According to WALS, the overwhelming majority or languages have SV order, so it seems like SV/SVO may be the way to go. Time to stare blankly at a ceiling for a while until I settle on either route.

Hah, I know the feeling well…

[Vardelm]

Devani: VS / OVS order

I would say I'm 98.5% certain I'm using VS / OVS order for the comment portion of Devani sentences. I picture topics working similar to what I understand of Japanese, where a topic can be introduced and following sentences might use non-topic particles to continue commenting. Devani will not have non-topic particles; the arguments will just be unmarked and no topic particle will appear in the sentence. Similar to Japanese, subjects and agents can be dropped, particularly if they are a pronoun that refers to the topic (such as my 1st example above). Even patients might be dropped, but probably only when they are the topic, and then the agent would need to be specified.

mechani the paramam mida mechani scholar the TOP param-am read-3P.VOL.VIS mida he.3P.ANIM Regarding the scholar: he is reading.

samhani the ne malkusuvam tujari samhani ceremony the TOP ne 2P.INF m-alkus-uv-am 3P-anoint.IRS-TRS.VOL-3P.VOL.VIS tujari priest For the ceremony, the priest will anoint you.

Some comments & reasons for this:

• One of the main reasons is that Devani uses postpositions, so OV order fits better.

• Added to the above, Devani will use stative verbs for adjectives, and this will include the "postpositions", which will just be verbs meaning "to be at", "to be under", etc. Devani "postpositions" will just be the participle of the corresponding stative verb.

• I already use SVO order and accusative(-ish) alignment in my other languages, so this adds a bit more variation.

• I picture Devani having developed from an SOV language, perhaps a marked accusative language, where the nominative marking turned into a topic marker and the subject was then relegated to coming after the verb, perhaps as an adjunct. This would explain Devani's head-final order & use of postpositions.

• I can see a bunch of directions for descendent languages to develop, either (back?) to a marked accusative lang, a non-topic prominent nom-acc language, & more.

• OVS can help push a future "Tibetan Dwarvish" to be OVS as well.

• The object volition markers and the subject volition/evidential/person markers would be in the same order as the arguments. Not a requirement at all, but a nice consistency in my mind.

• Topic - Verb - Agent word order, where the patient would have been a pronoun that was dropped and the topic serves as the patient, can provide a sort of passive voice since Devani doesn't (intentionally) make much use of voice.

[bradrn]

mechani the mida paramam mechani scholar the TOP paramam read mida he.3P.ANIM Regarding the scholar: he is reading.

samhain the ne malkusuvam tujari samhani ceremony the TOP ne 2P.INF m-alkus-uv-am 3P-anoint.IRS-TRS.VOL-3P.VOL.VIS tujari priest For the ceremony, the priest will anoint you.

I’ll just note that your glosses and sentences have gotten out of sync…

[Vardelm]

I’ll just note that your glosses and sentences have gotten out of sync…

Thanks as always for the editing services. Now fixed (I think!).