Is this attested? (Updated: comments appreciated!)

[dewrad]

So, since I was last active on here I've basically been rebuilding my conworld from the ground up (quite literally: I spent the best part of a month dicking around with GPlates simulating a new tectonic history for the entire planet). I've now got to the point that I'm working languages again: in this case it's the Nostratic-equivalent of the northern continent.

In an effort to think outside the box with regard to morphosyntactic alignment, I've come up with something that is kinda-sorta tripartite and kinda-sorta fluid-S, and I was wondering if it is attested in any natlangs.

Basically: the most fundamental opposition marked on nouns in this system is the degree of control that the subject has over the verb.

Nouns have three cases, which currently I'm labelling "ergative", "absolutive" and "oblique".

Verbs can be either stative (indicating, shockingly, a state- "adjectives" also fall in this category) or dynamic (indicating an action or a change of state). This is an aspectual distinction to a degree, but it is determined lexically rather than being a gammatical alternation per se.

Stative verbs are uniformly intransitive and, logically, the subject does not have control over the state. As such, the argument of the verb is marked with the absolutive case:

tana wetɬä

tan-

man-

a

ABS

wetɬä

sleeps

"The man sleeps."

For dynamic intransitive verbs, the subject is marked with the absolutive case if he lacks control over the action or the ergative if the subject has control. For example:

tanu rokʰä

tan-

man-

u

ERG

rokʰä

shouts

"The man shouts (on purpose, in order to warn his friend that a swarm of bees is heading in his direction for example)."

tana rokʰä

tan-

man-

a

ABS

rokʰä

shouts

"The man shouts (involuntarily, perhaps because he has been stung by a bee)."

Moving on to transitive verbs, the same marking obtains dependent on whether the subject is marked as [±control]. However, the patient of the verb is marked with the oblique case:

tana maɬi bätsʰä

tan-

man-

a

ABS

maɬ-

dog-

i

OBL

bätsʰä

kicks

"The man kicks the dog (accidentally)"

tanu maɬi bätsʰä

tan-

man-

u

ERG

maɬ-

dog-

i

OBL

bätsʰä

kicks

"The man kicks the dog (on purpose)"

For what it's worth, the oblique also marks the possessor in genitive constructions:

tani maɬu tani ŋʷaka

tan-

man-

i

OBL

maɬ-

dog-

u

ERG

tan-

man-

i

OBL

ŋʷaka

bites

"The man's dog bites the man."

Looking at it semantically, it appears that the agent maps to the ergative case, the experiencer to the absolutive and the patient/theme to the oblique. Is there a similar attested natlang? Off the top of my head, I can't think of one immediately, but my knowledge of languages beyond Europe and western Asia is pretty minimal.

[zompist]

Looking at it semantically, it appears that the agent maps to the ergative case, the experiencer to the absolutive and the patient/theme to the oblique. Is there a similar attested natlang?

That's pretty much the definition of a tripartite language. Nez Percé (Plateau, 'Penutian') is the example I've retained, but there are others.

And yeah, if degree of control is a big part of your grammar, you've probably got Fluid-S going. Tsova-Tush (NE Caucasian) specifically marks verbs by purpose.

Unfortunately, overall treatments of morphosyntax (e.g. Comrie, Payne or, um, me) tend to spend more of their attention on ergativity proper, so I couldn't tell you if, say, combining tripartite and fluid-S as you've done is attested. I don't see why not though.

[Kuchigakatai]

In terms of morphosyntactic alignments, isn't this language, technically speaking, a nominative-accusative language?

After all, S is never O here (that is, the subject of a monovalent intransitive verb ("S") is never marked as the object of a bivalent transitive verb ("O")), there is only S = A (that is, the subject of a monovalent intransitive verb ("S") is always marked in ways that the subject of a bivalent transitive verb ("A") is marked).

Of course, what actually happens in the conlang is much more interesting than that, since subjects can take an agentive case (ERG) or patientive case (ABS) depending on control, and the like, as is the case in true fluid-S languages. But the typology of morphosyntactic alignments is about how "S" appears in terms of marking compared to "A" and "O".

In true fluid-S languages, the marking of "S" would be the same as that of "O" when there is a lack of control, volition, etc. in the action of the verb. As in, say, [-control] "thinks me" (meaning 'I think', where S = O, supposing it were possible in English to have an impersonal verb like that, and assuming thinking is [-control]), which would contrast with [+control] "I eat" (where S = A).


To me, all this means is that the conventional typology is, naturally, not good enough to encompass a variety of unusual phenomena, like the one in this conlang.

[bradrn]

This sounds very realistic, although I would be surprised if it were attested. As zompist has said, this looks like a variant of a fluid-S system, but with the object marked with its own case rather than absolutive. In some ways, it reminds me of Sahaptin (as described by DeLancey here): in that language, S/A/O all get marked separately, but there are two case-markers for A (although the distinction is admittedly not related to agentivity).

One minor terminological note: since your ergative and your absolutive case are both used for S and A, I would call them both nominative instead. (Maybe ‘unagentive nominative’ and ‘agentive nominative’?) I also think that your ‘oblique’ case is actually an accusative case; as Ser has noted, this makes your language a peculiar variety of nominative-accusative.

… my knowledge of languages beyond Europe and western Asia is pretty minimal.

By the way, I’ve been writing a series of posts on ergativity; these cover areas which should be relevant to your question. You might find it useful in getting a better understanding of this area.

(Yes, I know this is self-promotion; hopefully it’s on-topic enough that no-one will mind… )

[akam chinjir]

I agree also with Ser, this looks more like a NOM/ACC pattern in which there are two cases available for S=A subjects. With just these examples you could relabel ERG as NOM, ABS as DAT, and OBL as ACC, and I think it'd all go through (and correspond to your semantic summary pretty well).

I actually don't know if there are any languages with fluid active/dative alignment, or as many dative subjects as you'd end up with on this interpretation. I was just reading about a pattern in West Circassian where some subjects get demoted to oblique in certain constructions (the case-marking doesn't change, because ergative uses the regular oblique case, but there's a difference in agreement). That allows one verb to vary, though it's not conditioned by control. Can't see why something like that couldn't be conditioned by control, though.

[bradrn]

I agree also with Ser, this looks more like a NOM/ACC pattern in which there are two cases available for S=A subjects. With just these examples you could relabel ERG as NOM, ABS as DAT, and OBL as ACC, and I think it'd all go through (and correspond to your semantic summary pretty well).

I’d be curious to know: why do you call dewrad’s ‘absolutive’ case dative? It gets used for both S and A, so surely it should be called nominative, if anything.

[akam chinjir]

I agree also with Ser, this looks more like a NOM/ACC pattern in which there are two cases available for S=A subjects. With just these examples you could relabel ERG as NOM, ABS as DAT, and OBL as ACC, and I think it'd all go through (and correspond to your semantic summary pretty well).

I’d be curious to know: why do you call dewrad’s ‘absolutive’ case dative? It gets used for both S and A, so surely it should be called nominative, if anything.

Because dative subjects are pretty widely attested, and dewrad seemed to be thinking of those arguments as experiencers, and dative subjects are especially widely attested for experiencers, and can be used for either S or A. Though whether that label makes sense would depend on what else these cases do in the language.

One thing: the relabeling of OBL as ACC means ACC is syncretic with GEN, which is a bit odd. (Okay, maybe it's really, really odd.)

Come to think of it, if you think of those as genuinely oblique arguments, maybe there are no transitive verbs at all, just unaccusative and unergative ones. In that case, dewrad's original case labels make perfect sense. I'm not sure if there are any languages with no surface-transitive verbs, but it's a cool idea.

[bradrn]

I agree also with Ser, this looks more like a NOM/ACC pattern in which there are two cases available for S=A subjects. With just these examples you could relabel ERG as NOM, ABS as DAT, and OBL as ACC, and I think it'd all go through (and correspond to your semantic summary pretty well).

I’d be curious to know: why do you call dewrad’s ‘absolutive’ case dative? It gets used for both S and A, so surely it should be called nominative, if anything.

Because dative subjects are pretty widely attested, and dewrad seemed to be thinking of those arguments as experiencers, and dative subjects are especially widely attested for experiencers, and can be used for either S or A. Though whether that label makes sense would depend on what else these cases do in the language.

Thanks for explaining! I had forgotten about dative subjects, but now that you mention those it makes perfect sense. (Although I hadn’t known that dative subjects are particularly common for experiencers; I’m sure I’ll find a use for that now that I know about it.)

One thing: the relabeling of OBL as ACC means ACC is syncretic with GEN, which is a bit odd. (Okay, maybe it's really, really odd.)

Why is this odd? It’s fairly well-known that ergative often functions as a locative, instrumental or genitive; similarly, I had assumed that accusative could also have these functions (since both ergative and accusative are marked cases).

Come to think of it, if you think of those as genuinely oblique arguments, maybe there are no transitive verbs at all, just unaccusative and unergative ones. In that case, dewrad's original case labels make perfect sense. I'm not sure if there are any languages with no surface-transitive verbs, but it's a cool idea.

I like this idea! Although admittedly this interpretation doesn’t seem very plausible…

[Kuchigakatai]

The accusative case serving for both direct objects and (bare) possessors is attested in Old French, especially in the earlier centuries. And a lot of nouns in Standard Arabic have identical accusative and genitive forms, more so in the plural than in the singular, and as a mandatory rule in the dual.

So the accusative case serving as a genitive didn't strike me as odd at all...

[akam chinjir]

Partially responding to both bradrn and Ser: I can't remember details, but I'm pretty sure Baker, in his book on case, reports that ACC/GEN syncretism is rare enough that he thinks this warrants one kind of explanation. I can't remember if he mentions any cases where it does happen. He certainly wouldn't rule out the case where accusative and genitive get conflated due to sound change. Part of the story's presumably diachronic. Diachronically, it's not hard to see how ergative cases can derive from genitive via nominalisations (or instrumental via passives), I don't know a similar story about accusative and genitive, though of course that could just be my imagination failing. (Also I could be misremembering or Baker could have things wrong, of course.)

[bradrn]

Partially responding to both bradrn and Ser: I can't remember details, but I'm pretty sure Baker, in his book on case, reports that ACC/GEN syncretism is rare enough that he thinks this warrants one kind of explanation. I can't remember if he mentions any cases where it does happen. He certainly wouldn't rule out the case where accusative and genitive get conflated due to sound change.

I think that this is the second time you’ve mentioned Baker on case; I’ll have to see if I can find a copy to read!

EDIT: Are you talking about this book? (Case: Its Principles and its Parameters) I do hope so, because I should be able to access it via my university library. (I say ‘should’ because when I try to log in it gives me an error…)

Part of the story's presumably diachronic. Diachronically, it's not hard to see how ergative cases can derive from genitive via nominalisations (or instrumental via passives)

I’m not sure how you could get ergative from genitive via nominalisation — could you elaborate? As for instrumental to ergative, I don’t think I’ve heard of that happening via passives; Dixon writes that it could come from sentences such as John opened the door with the key vs The key opened the door, in which the key is instrumental in the first sentence but can be interpreted as the transitive subject in the second.

[akam chinjir]

I’m not sure how you could get ergative from genitive via nominalisation — could you elaborate? As for instrumental to ergative, I don’t think I’ve heard of that happening via passives; Dixon writes that it could come from sentences such as John opened the door with the key vs The key opened the door, in which the key is instrumental in the first sentence but can be interpreted as the transitive subject in the second.

My memory's not supplying me with specific examples, unfortunately. It's pretty widely thought that ergative clauses can derive from passives, and part of that story is that the agent phrase in the passive gets reinterpreted as subject, but retaining whatever oblique morphology agents of passives typically got, which is often instrumental. Meanwhile I think it's supposed to be pretty common for nominalised clauses have ergative case patterns involving genitive, and also for subordinate clauses to be nominalised. I'm thinking here mostly about ergativity splits in which only subordinate clauses are ergative (or do I have that backwards?), though you can also imagine patterns that originate in subordinate clauses generalising to main clauses, I suppose.

(In Dyirbal in particular there's some reason to think that in transitive sentences it's the P argument that's in the position of the structural subject, with the A argument taking what's otherwise an oblique case marker, something like a passive even in synchronic grammar.)

[bradrn]

I’m not sure how you could get ergative from genitive via nominalisation — could you elaborate? As for instrumental to ergative, I don’t think I’ve heard of that happening via passives; Dixon writes that it could come from sentences such as John opened the door with the key vs The key opened the door, in which the key is instrumental in the first sentence but can be interpreted as the transitive subject in the second.

My memory's not supplying me with specific examples, unfortunately. It's pretty widely thought that ergative clauses can derive from passives, and part of that story is that the agent phrase in the passive gets reinterpreted as subject, but retaining whatever oblique morphology agents of passives typically got, which is often instrumental.

I did know that ergative can derive from passives; I just didn’t make the connection that the former subject of the passivised sentence could get instrumental case. (I’m more familiar with it being dative.)

Meanwhile I think it's supposed to be pretty common for nominalised clauses have ergative case patterns involving genitive, and also for subordinate clauses to be nominalised. I'm thinking here mostly about ergativity splits in which only subordinate clauses are ergative (or do I have that backwards?), though you can also imagine patterns that originate in subordinate clauses generalising to main clauses, I suppose.

That argument makes sense. But according to Dixon, there don’t seem to be terribly many patterns with main/subordinate ergativity splits: on semantic grounds, some subordinate clauses should be accusative (e.g. purposives) while others should be accusative. But Shokleng is an attested example of the sort of split you talk about, in which main clauses are accusative or ergative while subordinate clauses are ergative (although the split in Shokleng is pretty weird in several ways).

(In Dyirbal in particular there's some reason to think that in transitive sentences it's the P argument that's in the position of the structural subject, with the A argument taking what's otherwise an oblique case marker, something like a passive even in synchronic grammar.)

This sounds like a pretty weird argument — what is the reasoning behind it?

[akam chinjir]

(In Dyirbal in particular there's some reason to think that in transitive sentences it's the P argument that's in the position of the structural subject, with the A argument taking what's otherwise an oblique case marker, something like a passive even in synchronic grammar.)

This sounds like a pretty weird argument — what is the reasoning behind it?

I'm pretty sure it derives from the classic Comrie/Keenan paper on relative clauses, fwiw. The basic idea is that the reason why the ABS argument ends up with properties you'd normally associate with a subject is that it actually is a subject. I've only just started reading seriously about this sort of analysis of syntactically ergative languages, but it seems to have been applied to a fairly broad range of languages.

(For some purposes you still have to count the ergative argument as a subject---so these approaches tend to agree with the idea that the concept of a subject sort of comes apart in syntactically ergative languages. But in the formal approaches to syntax that I'm thinking of the concept of a subject tends not to have any deep significance in the first place.)

[Pabappa]

"The boy's apple" / "the boy eats his apple"

Later becomes
"The boy's eats (his/the ) apple " and genitive is now ergative

[Kuchigakatai]

Partially responding to both bradrn and Ser: I can't remember details, but I'm pretty sure Baker, in his book on case, reports that ACC/GEN syncretism is rare enough that he thinks this warrants one kind of explanation. I can't remember if he mentions any cases where it does happen. He certainly wouldn't rule out the case where accusative and genitive get conflated due to sound change. Part of the story's presumably diachronic. Diachronically, it's not hard to see how ergative cases can derive from genitive via nominalisations (or instrumental via passives), I don't know a similar story about accusative and genitive, though of course that could just be my imagination failing. (Also I could be misremembering or Baker could have things wrong, of course.)

I just checked nominal declension in Akkadian, and in that language, the dual and plural have mandatory accusative-genitive syncretism as a rule (unlike Standard Arabic, where the plural may still distinguish nom., acc. and gen. depending on the type of plural stem used).

I also checked my go-to book on Old Occitan (William D. Paden's intro tome), and, as in Old French (la terre mes enfanz 'the land (of) my children'), the accusative also has genitive function there (la terra mos enfans).

I don't know what Mark Baker says in what you read, but maybe he skipped medieval Western Romance and historical Semitic in his data, and funnily it turns out those are most of the languages with case I happen to be somewhat familiar with...

At least I can say Classical and Late Latin, as written, really don't have any accusative-genitive syncretism, not even in irregular nouns (unless you count undeclinable foreign proper names or something). Well, there's the thing about the Latin 4th masculine/feminine declension with syncretic genitive singular and accusative plural (huius cēnsūs, hōs cēnsūs), and some Attic Greek nouns ending in /a/ or /a:/ similarly have identical gen. sg. and acc. pl. (tês theōríās, tâs theōríās), but I'm not counting these since the number is different...

[Moose-tache]

Since there's a lot of discussion about the proper labeling of things ("Is this absolutive? Is it dative? Is it nominative?") I thought I would point out a fun example: in Choctaw, they just don't bother to name the agreement affixes. Instead you have Class I (mostly agents), Class II (sometimes agent, sometimes experiencer or other low-agency subject, often direct object, also related to possessive markers), and Class III (mostly indirect objects, sometimes direct objects, or rarely even experiencers). Usually if a verb takes a Class I subject it will take a Class II object, and if it takes a Class II subject it will take a Class III object. This is mostly but not entirely lexically determined. The whole thing would be a nightmare to classify consistently, so they just didn't. Pretty smart, in my opinion. In that vein, you could call your cases U, A, and I, and just describe their usage. Leave it to the philosophers whether u is named "agentive" or "ergative" or "agento-nominative" or whatever.

[akam chinjir]

I don't know what Mark Baker says in what you read, but maybe he skipped medieval Western Romance and historical Semitic in his data, and funnily it turns out those are the most of the languages with case I happen to be somewhat familiar with...

He's got a bit of a balancing act---for the things he wants to investigate, he needs to go pretty deeply, he can't just depend on taking a look at a few pages in a grammar. So in that book he ends up picking about 20 languages that he studies with some care (Amharic and Tamil are on the list, but no other Semitic or Indo-European). (Coincidentally, Choctaw is also on the list.)

[dewrad]

Thanks for all the replies guys!

I agree also with Ser, this looks more like a NOM/ACC pattern in which there are two cases available for S=A subjects. With just these examples you could relabel ERG as NOM, ABS as DAT, and OBL as ACC, and I think it'd all go through (and correspond to your semantic summary pretty well).

While I get where you're coming from, I'm loath to relabel ABS as DAT, primarily becase the case lacks any dative functions. Indirect objects (along with other non-core participants) are marked with postpositions, which govern the oblique case:

tanu maɬi-na pʰetre yotsʰä

tan-

man-

u

ERG

maɬ-

dog-

i

OBL

na

to

pʰetr-

bone-

i

OBL

yotsʰä

gives

"The man gives a bone to the dog."

It's also used to mark the relationship of modifier to head in noun-noun compounds:

demekʰotnä

dem-

winter-

i-

OBL-

kʰotnä

campsite

"Winter dwelling"

From a diachronic point of view (for when do I not look at everything in this way?), the instant system is supposed to be the descendant of a relatively straightforward ergative-absolutive system. However, with speakers coming to use the ergative and absolutive to mark control, patients of transitive verbs came to be kind of compounded with their verbs - we could see the use of the oblique case to mark the patient of transitive verbs as in fact a species of light object incorporation maybe?

Come to think of it, if you think of those as genuinely oblique arguments, maybe there are no transitive verbs at all, just unaccusative and unergative ones. In that case, dewrad's original case labels make perfect sense. I'm not sure if there are any languages with no surface-transitive verbs, but it's a cool idea.

I like this idea! Although admittedly this interpretation doesn’t seem very plausible…

So do I! In fact, I'm tempted to run with this. The humans of my conworld are not supposed to be Homo sapiens, but a closely related (sub-)species like Neanderthals or Denisovans, so I like the idea of having a kind of marking which is plausible but absent in terrestrial languages.

As an aside, one of the things I like the most about this system is the potential for turning into different alignments in the daughter languages.

[Richard W]

My memory's not supplying me with specific examples, unfortunately. It's pretty widely thought that ergative clauses can derive from passives, and part of that story is that the agent phrase in the passive gets reinterpreted as subject, but retaining whatever oblique morphology agents of passives typically got, which is often instrumental.

I did know that ergative can derive from passives; I just didn’t make the connection that the former subject of the passivised sentence could get instrumental case. (I’m more familiar with it being dative.)

It's quite common in Pali to find formally passive past tense sentences that are best translated as active sentences. I understand this also shows up in Sanskrit.