Conlang Random Thread

[Ahzoh]

The only thing I would criticize about your table there is that I'd call what you call 'intransitive' 'direct' instead.

It's the Official(tm) terminology for the S case in tripartite alignment, and Ergative for the A and Accusative for the P/O.

Thing is, you don't have tripartite alignment. I'm not sure what the name of the alignment is, but there's got to be a name somewhere. (I personally like this alignment and have a tendency to use it, alongside direct-inverse marking, in my languages.)

Well no, it's split ergative, but I have the table of cases arranged in this fashion to show the diachronic origins of the cases from an older tripartitie paradigm. In animates, the intransitive case merges with the ergative resulting in nom-acc alignment. In inanimates, the intransitive case merges with the accusative resulting in erg-abs alignment.

The aesthetics of the presentation of information is more important to me than adhering to standard convention, but I also want to have my cake and eat it too with regards to adhering to standard conventions.

[Ahzoh]

I didn't think it was a problem to have a 4-case (nom, acc, abs, erg) analysis of split-erg. In fact, I thought that was pretty standard, but apparently it is not. So now I've wasted days thinking about how to have a nice-looking table that satisfies a 3-case (nom, acc, erg) analysis.

People with some information about cases and what they mean should be able to look at my table at a glance and immediately understand the morphosyntactic behavior of a particular syntactic case. Without needing any special blurbs of information.

[bradrn]

I'm not sure what the name of the alignment is, but there's got to be a name somewhere. (I personally like this alignment and have a tendency to use it, alongside direct-inverse marking, in my languages.)

I’m tempted to call it the ‘Conlanger Inverse System’…

If I call the absolutive an accusative it would have to be on the same row as the animate accusatives, but I would also have to add a nominative row and merge the cells to convey syncretism between the nominative and accusative, like in here:
https://en.wikipedia.org/wiki/Hittite_language#Nouns

Given what you say here, I tried to look at the table, and I think I’ve gotten myself confused yet again. (This is the problem with poor terminology!)

So, let me first confirm that my understanding of your second table is correct:

• Take S to be the sole object of an intransitive verb, A the subject of a transitive verb, O the object of a transitive verb.
• The case -am etc., labelled ‘nominative’ for animate nouns, marks S and A for animate nouns.
• The case -as etc., labelled ‘accusative’ for animate nouns and ‘absolutive’ for inanimate nouns, marks O for all nouns, extended to S for inanimate nouns.
• The case -an etc., labelled ‘instrumental’ for animate nouns and ‘ergative’ for inanimate nouns, marks peripheral NPs, as well as A for inanimate nouns.

Assuming this is correct: it really is quite confusing. You’ve taken what should be ‘a single case’ — in that it has the same form for animate and inanimate nouns, and closely related functions for both — and split it up into two separate rows of the table. That only obscures the situation.

Instead, I would suggest the following:

• Call -am the ‘nominative’, making clear in the table that it’s only available for animate nouns.
• Call -as the ‘absolutive’ across all nouns, because it’s used for S and O arguments. (The fact that you have both a ‘nominative’ and ‘absolutive’ case implies there is a clash for S: in your case the clash is resolved via animacy.)
• Call -an the ‘ergative’, because IIRC that’s the usual terminological choice for a syncretic ergative/instrumental case. (Though honestly I’d be fine with ‘ergative/instrumental’ here — it’s more obviously syncretic than the other cases.)

This lets you put them quite neatly in a table, like so:

vrkhazian-cases.png (21.98 KiB) Viewed 2448 times

(I didn’t bother writing out the whole thing, but this should be sufficient to show the idea. I decided not to centre the vocative row, to highlight its relationship to the other two cases.)

I didn't think it was a problem to have a 4-case (nom, acc, abs, erg) analysis of split-erg. In fact, I thought that was pretty standard, but apparently it is not. So now I've wasted days thinking about how to have a nice-looking table that satisfies a 3-case (nom, acc, erg) analysis.

I mean, no-one’s saying that having four names is an invalid analysis, as such. It’s just not a parsimonious one, in that you can describe it just as well with three names. And I would argue that the three-name analysis reveals the underlying structure of your system better than the four-name one does.

People with some information about cases and what they mean should be able to look at my table at a glance and immediately understand the morphosyntactic behavior of a particular syntactic case. Without needing any special blurbs of information.

This seems to be a common misconception amongst conlangers, that a linguistic term uniquely and completely defines the behaviour of a morpheme. Actually, it seems quite common even amongst actual linguists — you’ll see grammars with a morpheme listed only as ‘nominative’ or ‘perfective’ or ‘intensive’ or so on, without any further elaboration.

I think this is utter rubbish. It is foolish to expect markers to behave in precisely the same way across languages — no matter how similar said markers may seem to each other. This kind of terminological misconception caused me a lot of trouble when I was trying to learn linguistics: indeed, I’d go so far as to say that eradicating this misguided belief has been key to most of my understanding of linguistics.

So, it’s simply impossible to avoid ‘any special blurbs of information’, if you want to make your language naturalistic. You’ll need long-form descriptions of each and every case anyway, because the terms are just too ambiguous on their own. Trying to achieve the goal of ‘immediate understanding at a glance’ would be like nailing jelly to a wall (as they say). So you might as well choose terms which accord with the most parsimonious analysis of your language, rather than trying desperately to find the nonexistent terms which precisely match the details of the behaviour in this specific language.

[Ahzoh]

I'm not sure what the name of the alignment is, but there's got to be a name somewhere. (I personally like this alignment and have a tendency to use it, alongside direct-inverse marking, in my languages.)

I’m tempted to call it the ‘Conlanger Inverse System’…

It's split-ergative. The only difference is that the accusative and absolutive share a morpheme instead of the nominative and absolutive sharing a morpheme.

Although, maybe that means it's what they call the "split marked-S" alignment, if one considers the -as case the least marked case.

There is also this interesting quote on the Ergativity for Novices Thread:

Some linguists argue that systems with NP splits should be understood as tripartite, but with syncretism, e.g. ERG/NOM syncretism in pronouns and NOM/ACC syncretism in common nouns. Dyirbal has an interesting data point to support this: it has case concord, and pronouns can take modifiers, and when (say) a first person pronoun is the subject of a transitive verb, it may look nominative (=unmarked), but any modifier will show ergative case.

Only my split is between animates and inanimates and not between full nouns and pronouns.

[bradrn]

It's split-ergative. The only difference is that the accusative and absolutive share a morpheme instead of the nominative and absolutive sharing a morpheme.

Firstly, ‘split-ergative’ is a dreadfully imprecise term covering many different alignment systems. Secondly, there’s also the direct–inverse subsystem to consider.

Also, having the syncretism be accusative/absolutive rather than nominative/absolutive is a bigger difference than you might think… it means the split is on the intransitive argument, not just the transitive arguments.

Although, maybe that means it's what they call the "split marked-S" alignment, if one considers the -as case the least marked case.

This characterisation sounds more accurate to me. Of course, you do need to justify why you consider -as the least marked case.

There is also this interesting quote on the Ergativity for Novices Thread:

Some linguists argue that systems with NP splits should be understood as tripartite, but with syncretism, e.g. ERG/NOM syncretism in pronouns and NOM/ACC syncretism in common nouns. Dyirbal has an interesting data point to support this: it has case concord, and pronouns can take modifiers, and when (say) a first person pronoun is the subject of a transitive verb, it may look nominative (=unmarked), but any modifier will show ergative case.

Only my split is between animates and inanimates and not between full nouns and pronouns.

Well, you said your system evolved from a tripartite one, right? So this matches the evolution. It’s also how I presented it in my table, with the vocative (which I presume was the original intransitive case) variously merged with the nominative or absolutive.

[Ahzoh]

Firstly, ‘split-ergative’ is a dreadfully imprecise term covering many different alignment systems. Secondly, there’s also the direct–inverse subsystem to consider.

imprecise to people who spend all day splitting hairs.

For me and possibly lay people, the term is enough to convey "syntactic or morphological erg-abs behavior in one situation and nom-acc behavior in another".

Also, having the syncretism be accusative/absolutive rather than nominative/absolutive is a bigger difference than you might think… it means the split is on the intransitive argument, not just the transitive arguments.

In terms of real-life (out-of-universe) conlanging development, it started as a normal nom-acc alignment across the board, with a syncretized nominative and accusative case in inanimate nouns. But then I thought it would be neat if the instrumental case marker -n also marked inanimate agents of transitive verbs.

I also have the constraints that -m and -n cannot coexist in the nom and acc/abs and erg cases (e.g. only -m vs -s or -n vs -s but not -m vs -n) because I consider the distinction between nasals word-finally to be too weak to create a meaningful and stable contrast.
At the same time I also didn't want nominative -s with accusative -m (too Latinate).

Also should note the instrumental marks the theme of a ditransitives verbs, analogous to how a dative marks the recipient. So I'm not sure how peripheral those roles are wrt ditransitives. So I tend to consider datives and instrumentals as core cases equal to nominatives and accusatives.

Although, maybe that means it's what they call the "split marked-S" alignment, if one considers the -as case the least marked case.

This characterisation sounds more accurate to me. Of course, you do need to justify why you consider -as the least marked case.

Considering all the cases equally involve a vowel and a consonant, I find it hard to think any of the cases are "least marked" although I consider the -Vm case to be the citation form of animates and the -Vs case to be the citation form of inanimates.

Well, you said your system evolved from a tripartite one, right? So this matches the evolution. It’s also how I presented it in my table, with the vocative (which I presume was the original intransitive case) variously merged with the nominative or absolutive.

The Vocative was originally conceived as a "second nominative" or a specialized variant of the nominative case. So maybe it originated as a specialized intransitive case. Maybe its form is a relic of the nominal system before the case system developed.
Cuz before the case system existed, it was just the gender endings and no case marking.

[Travis B.]

On second thought, this system seems different from mine in that my direct case is unmarked, being the bare stem without any ending, and both my ergative and accusative cases are explicitly marked by their very nature (hence why I named them such), whereas in Vrkhazhian there are really are no unmarked cases in the absolute state.

[Ahzoh]

I come up with these examples to illustrate what's all going on:

mouse-m runs
mouse-m bites man-s
mouse-m bites cheese-s

pillar-s falls
pillar-n crushes man-s
pillar-n crushes cheese-s

This is different than the apparently typical way:

mouse-s runs
mouse-s bites man-m
mouse-s bites cheese-s

pillar-s falls
pillar-n crushes man-m
pillar-n crushes cheese-s

[keenir]

Firstly, ‘split-ergative’ is a dreadfully imprecise term covering many different alignment systems. Secondly, there’s also the direct–inverse subsystem to consider.

imprecise to people who spend all day splitting hairs.

For me and possibly lay people, the term is enough to convey "syntactic or morphological erg-abs behavior in one situation and nom-acc behavior in another".

your target audience is both laypeople and those who spend all day splitting hairs? and here I thought Tantalaus had a rough task in Hades.

very owch!

but i wish you good luck in pursuiing it.

[Ahzoh]

your target audience is both laypeople and those who spend all day splitting hairs? and here I thought Tantalaus had a rough task in Hades.

very owch!

but i wish you good luck in pursuiing it.

I don't have a strictly defined target audience.

[bradrn]

Firstly, ‘split-ergative’ is a dreadfully imprecise term covering many different alignment systems. Secondly, there’s also the direct–inverse subsystem to consider.

imprecise to people who spend all day splitting hairs.

For me and possibly lay people, the term is enough to convey "syntactic or morphological erg-abs behavior in one situation and nom-acc behavior in another".

(Leaving aside the fact that most laypeople have never heard of split ergativity…)

This is, indeed, what that term conveys. Doesn’t change the fact that it’s imprecise — it covers a huge range of very different alignment systems. (Dyirbal, Kala Lagaw Ya, Hindi, Mayan, Georgian…) Saying ‘this language has split ergativity’ is OK, but it’s far from a complete description.

Although, maybe that means it's what they call the "split marked-S" alignment, if one considers the -as case the least marked case.

This characterisation sounds more accurate to me. Of course, you do need to justify why you consider -as the least marked case.

Considering all the cases equally involve a vowel and a consonant, I find it hard to think any of the cases are "least marked" although I consider the -Vm case to be the citation form of animates and the -Vs case to be the citation form of inanimates.

There’s formal markedness, and then there’s functional markedness. The latter covers citation form etc. From this description, it sounds like no one case can be considered ‘least marked’.

This is different than the apparently typical way:

mouse-s runs
mouse-s bites man-m
mouse-s bites cheese-s

pillar-s falls
pillar-n crushes man-m
pillar-n crushes cheese-s

I’d go a bit further here, and say that the ‘typical’ animacy-based split looks like this:

mouse-∅ runs
mouse-∅ bites man-ACC
mouse-∅ bites cheese-∅

pillar-∅ falls
piller-ERG crushes man-ACC
pillar-ERG crushes cheese-∅

The idea therefore being that nouns are unmarked when they appear in their expected position, and marked when they appear in a position which contradicts their animacy. The basic idea is really ‘marked vs unmarked’, not ‘ergative vs absolutive’. (You can even merge the accusative and ergative cases, as in e.g. Proto-Pamir, but that’s an unstable system.)

[Ahzoh]

(Leaving aside the fact that most laypeople have never heard of split ergativity…)

Yeah... I'd have to explain it, but at least they wouldn't then proceed to "um, ackshually" me.

There’s formal markedness, and then there’s functional markedness.

I wonder how they determine "least marked" in languages without null marking, like, say, Akkadian, where nom is -u, -acc is -a and gen is -i.

I’d go a bit further here, and say that the ‘typical’ animacy-based split looks like this:

mouse-∅ runs
mouse-∅ bites man-ACC
mouse-∅ bites cheese-∅

pillar-∅ falls
piller-ERG crushes man-ACC
pillar-ERG crushes cheese-∅

The idea therefore being that nouns are unmarked when they appear in their expected position, and marked when they appear in a position which contradicts their animacy. The basic idea is really ‘marked vs unmarked’, not ‘ergative vs absolutive’. (You can even merge the accusative and ergative cases, as in e.g. Proto-Pamir, but that’s an unstable system.)

Looking back at the table, since one can apparently use the term "ergative" to denote peripheral NPs, I guess I might as well have nom and acc and merge the cells for the inanimate side.

---

I had a thought that the independent subject-marking pronouns shall become more like optional emphatic or topic markers. They would presumably be the topic head of a topic phrase while the rest of the sentence is a comment. This would be the only situation where the subject occurs before the verb, which is otherwise always the first element of a clause.

[Travis B.]

I have a number of verbs such as icha 'please' which have the experiencer as a patient, this results in an awkward situation where (typically) animate experiencers are marked when they would actually be unmarked given that, well, experiencers tend to be animate. Yet at the same time, I do not want to make the experiencer the agent, as, after all the patient is made to experience something (and especially since there are case where both an animate agent and an animate experiencer* make sense, and swapping them would not make sense).

I think the best solution is to make a class of verbs where both the agent and patient are agreed-with and case-marked with marked inanimates (like agents for normal verbs). I think I won't touch inverse marking, though.

* For instance:

• bed, yetti

Ichac'anumam da yetti
please-LOC-HAB-PAT.1.S-INV PROX.S.ANIM bed
He/she pleases me in bed.

[Travis B.]

I’d go a bit further here, and say that the ‘typical’ animacy-based split looks like this:

mouse-∅ runs
mouse-∅ bites man-ACC
mouse-∅ bites cheese-∅

pillar-∅ falls
piller-ERG crushes man-ACC
pillar-ERG crushes cheese-∅

The idea therefore being that nouns are unmarked when they appear in their expected position, and marked when they appear in a position which contradicts their animacy. The basic idea is really ‘marked vs unmarked’, not ‘ergative vs absolutive’. (You can even merge the accusative and ergative cases, as in e.g. Proto-Pamir, but that’s an unstable system.)

That's exactly what I envisioned for my language, except that I combined it with fluid-S. so one also gets:

pillar-ERG runs
mouse-ACC falls

[Ahzoh]

I think I can see why my case behaviour is so strange/confusing:

Original tripartite:
S = -m (intransitive case)
O = -s (accusative case)
A = -n (ergative case)

Three possibilities for mergers/syncretism that result in an alignment split:

Circular (bidirectional? syncretism)
S overtakes A, and both have -m
O overtakes S, and both have -s

Extension (unidirectional syncretism):
S overtakes A, and both have -m
S overtakes O, and both have -m

Inverse Extension (unidirectional syncretism):
A overtakes S, and both have -n
O overtakes S, and both have -s

Vrkhazhian is basically the first set.

[bradrn]

There’s formal markedness, and then there’s functional markedness.

I wonder how they determine "least marked" in languages without null marking, like, say, Akkadian, where nom is -u, -acc is -a and gen is -i.

It comes down to, essentially: which case is chosen when none in particular applies? Typically this would include situations like citation form, occurrence with an adposition, etc., although the precise criteria are language-dependent.

Looking back at the table, since one can apparently use the term "ergative" to denote peripheral NPs, I guess I might as well have nom and acc and merge the cells for the inanimate side.

Since your terminological choices are still confusing me a bit, could you show me what this resulting table would look like?

The idea therefore being that nouns are unmarked when they appear in their expected position, and marked when they appear in a position which contradicts their animacy. The basic idea is really ‘marked vs unmarked’, not ‘ergative vs absolutive’. (You can even merge the accusative and ergative cases, as in e.g. Proto-Pamir, but that’s an unstable system.)

That's exactly what I envisioned for my language, except that I combined it with fluid-S. so one also gets:

pillar-ERG runs
mouse-ACC falls

I’d hardly call this ‘exactly’ the same… like I said, I consider the change from unmarked-S to split-intransitive to be a pretty significant difference. (Not least because it has repercussions throughout the grammar: e.g., it seems to prevent any kind of syntactic ergativity.)

Also, while I’m complaining about terminology, what you have here is particularly confusing, because it means both ‘ERG’ and ‘ACC’ cases can apply to S. I’d instead call them ‘nominative’ and ‘absolutive’, respectively.

I have a number of verbs such as icha 'please' which have the experiencer as a patient, this results in an awkward situation where (typically) animate experiencers are marked when they would actually be unmarked given that, well, experiencers tend to be animate. Yet at the same time, I do not want to make the experiencer the agent, as, after all the patient is made to experience something (and especially since there are case where both an animate agent and an animate experiencer* make sense, and swapping them would not make sense).

It’s worth noting that these verbs often get odd case-marking. Such constructions as dative subjects are well-known for experiencers.

(There’s a nice paper by Georg Bossong surveying precisely these kinds of verbs in the languages of Europe. Alas, it is in French, so I’ve been struggling to read it…)

[Ahzoh]

Since your terminological choices are still confusing me a bit, could you show me what this resulting table would look like?

Exactly the table you made, but instead of the - on the nominative row on the inanimate side, it is merged with the vocative and absolutive.

The difficulty in terminological determination is probably compounded because of what I described in the above post I made.

It comes down to, essentially: which case is chosen when none in particular applies? Typically this would include situations like citation form, occurrence with an adposition, etc., although the precise criteria are language-dependent.

That is tricky, the object of adposition is majority of the time governed by the genitive case except for the special prepositions sa and ma which convey extra case roles (such as allative and comitative) and have different meanings depending on if the noun is in the accusative, genitive, or instrumental.

But the citation form of animates is the -m case and for inanimates it is the -s case.

---

I have devised every permutation of the animate/inanimate interactions

Animate subject/agent:
1.a) run-3sg mouse-NOM (VS)
1.b.i) bite-3sg mouse-NOM man-ACC (VSO)
1.b.ii) bite-3sg man-ACC mouse-NOM (VOS)
1.c) bite-3sg mouse-NOM cheese-ACC (VSO)
1.d) bite-3sg man-ACC (VO)
1.e) bite-3sg cheese-ACC (VO)

Inanimate subject/agent:
2.a) fall-3sg pillar-ACC (VS)
2.b) crush-3sg man-ACC pillar-ERG (VOS)
2.c.i) crush-3sg pillar-ERG cheese-ACC (VSO)
2.c.ii) crush-3sg cheese-ACC pillar-ERG (VOS)
2.d) crush-3sg man-ACC (VO)
2.e) crush-3sg cheese-ACC (VO)

Debating if verbs should have ergative agreement when inanimate agents are involved, just like in Hindi. This is relevant for matters of subject/object dropping.

[Travis B.]

The idea therefore being that nouns are unmarked when they appear in their expected position, and marked when they appear in a position which contradicts their animacy. The basic idea is really ‘marked vs unmarked’, not ‘ergative vs absolutive’. (You can even merge the accusative and ergative cases, as in e.g. Proto-Pamir, but that’s an unstable system.)

That's exactly what I envisioned for my language, except that I combined it with fluid-S. so one also gets:

pillar-ERG runs
mouse-ACC falls

I’d hardly call this ‘exactly’ the same… like I said, I consider the change from unmarked-S to split-intransitive to be a pretty significant difference. (Not least because it has repercussions throughout the grammar: e.g., it seems to prevent any kind of syntactic ergativity.)

Also, while I’m complaining about terminology, what you have here is particularly confusing, because it means both ‘ERG’ and ‘ACC’ cases can apply to S. I’d instead call them ‘nominative’ and ‘absolutive’, respectively.

The reason why I call them 'ergative' and 'accusative' are that they are marked, while canonically 'nominative' and 'absolutive' are used with unmarked arguments. Unmarked-S is direct (and has a bare nominal stem), for both animate agentive S and inanimate patientive S. 'Run' normally expects an agentive argument (someone being forced to run could be a patientive argument, though) while 'fall' normally expects a patientive argument (unless the intransitive argument intentionally fell, where then it would take an agentive argument. This results in things like (where 'mouse', being animate, is unmarked when agentive and marked when patientive):

mouse-∅ runs (the mouse ran normally, here 'mouse' is agentive)
mouse-ACC runs (the mouse was chased, e.g. by a cat; here 'mouse' is patientive)

mouse-ACC falls (the mouse unintentionally fell, here 'mouse' is patientive)
mouse-∅ falls (the mouse jumped, here 'mouse' is agentive)

Similarly:

pillar-ERG falls (somehow the pillar has a will of its own, and decided to fall, here 'pillar' is agentive)

So yeah, all three core cases, 'direct', 'ergative', and 'accusative' can be used with S, in addition to 'direct' and 'ergative' being used with A and 'direct' and 'accusative' being used with O. The easiest way to think about this is that unmarked arguments are always 'direct', while marked agentive (whether A or S) arguments are always 'ergative' and marked patientive (whether O or S) arguments are always 'accusative'.

I have a number of verbs such as icha 'please' which have the experiencer as a patient, this results in an awkward situation where (typically) animate experiencers are marked when they would actually be unmarked given that, well, experiencers tend to be animate. Yet at the same time, I do not want to make the experiencer the agent, as, after all the patient is made to experience something (and especially since there are case where both an animate agent and an animate experiencer* make sense, and swapping them would not make sense).

It’s worth noting that these verbs often get odd case-marking. Such constructions as dative subjects are well-known for experiencers.

(There’s a nice paper by Georg Bossong surveying precisely these kinds of verbs in the languages of Europe. Alas, it is in French, so I’ve been struggling to read it…)

With that in mind (I just thought of 'mir ist kalt' for instance), I decided to leave things as is, and treat the weirdly marked and agreed-with patients for experiencers as a feature rather than a bug.

[Glass Half Baked]

How many distinct registers do you all build into your conlangs? I've done up to three, but most often zero. Martin Joos believed there were five registers in English, but I've never worked out that many for a conlang in any detail.

[Travis B.]

How many distinct registers do you all build into your conlangs? I've done up to three, but most often zero. Martin Joos believed there were five registers in English, but I've never worked out that many for a conlang in any detail.

Honestly I haven't really done much of anything with register in my conlangs myself.