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

[Vardelm]

Very nice. I'd love to see the diachronics for these sometimes!

Thanks! Actually, the diachronics will probably never happen. These 4 languages I'm working on are intended as the PIEs of my world. I might have some clue about where they might have come from, but that's about it. Haven't thought that far ahead for the classes. I already have a version 3, which is fairly close to this one.

Is assignment and agreement going to be purely semantic? Or will some nouns be assigned to an unexpected noun class if they happen to begin/end with an appropriate affix?

That's a good suggestion. Probably pure semantic when I eventually get to the 1st run on vocab, but I might do a little with with stuff like this.

[Vardelm]

What is the upper limit/range on the number of inflections a language might have in (poly)personal agreement, and what languages are examples?

For Yokai, I'm planning on polypersonal agreement. The language is also intended as quite fusional, such that the agreement for subject & object would be 1 affix. However, I also have 25 noun classes which means that there would be well over 600 suffixes just for personal agreement. Based on what I've seen, many Bantu languages have noun classes & agreement/concords for both subject & object, but the subject & object are separate affixes (or perhaps clitics?). That seems like the memory load would be much lower since there would be 10-20 affixes for subject & object each. They are suffixed separately since the languages are agglutinating.

To me it seems like 600+ forms for 1 suffix "slot" would be pretty unstable and simplified quickly, but maybe I'm wrong on that.

Any thoughts and/or suggestions?

[Vilike]

That depends: fusion does not exclude regularity, especially if analogy works its magic along the line.
Given intransitive agreement markers ma-, ti-, sol- corresponding to three noun classes, the following combinations for transitive agreement could occur:

	ma-	ti-	sol-
ma-	mã-	mak-	maad-
ti-	tĩ-	tik-	tiid-
sol-	sũ-	suk-	suud-

What we see here is that ma- is always reduced to nasalisation, ti- to -k- and sol- to lengthening + -d-; additionally, the latter always has the form su- when before another formant.

But a larger degree of fusion could also mean that some combinations become homophonous. In Italian, for example, the six persons are distinguished from one another in all TAM, except in the subjunctive present, where all singular forms have the same ending, and the subjunctive imperfect, where 1sg and 2sg are the same: in these cases, pronouns are compulsory. You could also have disambiguating clitics somewhere on the verbal complex, if full pronouns are too much.
So given the same markers with a different outcome:

	ma-	ti-	sol-
ma-	mã-	mat-	mat-
ti-	tĩ-	tit-	tit-
sol-	sũ-	sut-	sut-

The last two columns could be distinguished with clitics -ti and -su respectively after the verbal complex.

Last possibility: if you are intent to keep fusion irregular, you could keep the number of combinations down with a direct-inverse system, so that there is only one affix for any combination of class markers and persons. Your noun classes look like they could lend themselves well to a hierarchy.
That's the Algonquin strategy to an extent (no really noun classes there, but different affixes and combinations of affixes depending on the animacy of the participants and TAM marking).
With our made-up markers again:

	ma-	ti-	sol-
ma-	mã-	mak-	maad-
ti-	mak-	tik-	tiid-
sol-	maad-	tik-	suud-

Here the hierarchy is ma- > ti- > sol-, more exactly the hierarchy between whatever class they're standing for.

[Vardelm]

Last possibility: if you are intent to keep fusion irregular, you could keep the number of combinations down with a direct-inverse system, so that there is only one affix for any combination of class markers and persons. Your noun classes look like they could lend themselves well to a hierarchy.

Since you have mentioned it, here's a spoiler:

A direct-inverse system is indeed my plan for Yokai and has been since my first concepts a few years ago. Further, as of yesterday, I am also strongly leaning towards one affix per combination of class markers - as you suggest - where which is the subject or object doesn't matter. At this point, I'd say that's definitely the direction I'll go.

With our made-up markers again:

	ma-	ti-	sol-
ma-	mã-	mak-	maad-
ti-	mak-	tik-	tiid-
sol-	maad-	tik-	suud-

Here the hierarchy is ma- > ti- > sol-, more exactly the hierarchy between whatever class they're standing for.

So, yeah, it would be something very similar to your chart, although larger due to the number of classes. The noun classes would only show up in 3rd person, so a bit of math for the total number of inflections:

1_{(1st person)} x 1_{(2nd person)} = 1
1_{(1st person)} x 25_{(3rd person classes)} = 25
1_{(2nd person)} x 25_{(3rd person classes)} = 25
Combinations (3rd person) = ₂₅C₂ = 300

Total = 351

That's significantly less than the 600+ I had posted above (based on just 25x25), but it's still a lot. Analogy & leveling might take care of quite a few, but I still have a hard time seeing that brought down to a manageable level.

I'm considering a slightly different tack. Instead of using the 25 "micro classes", I may just use the "macro classes", which are the prefix that indicates a plural noun. (See this post on Yokai noun classes. The letters are the "macro class" and number "micro class".) This would yield a much smaller number of inflections:

Combinations (1st, 2nd, + 5 macro-classes 3rd) = ₇C₂ = 21

Adjectives would agree with the micro-class of the noun since it would just be 25 due to no combination of 2 referents. Intransitive verbs could be either micro or macro agreement; not sure what I'd do with those yet.

This seems like it might also reinforce the use of both the prefixed macro-class & suffixed micro-class on nouns (since there would be agreement with both), making the whole system a more stable.

[bradrn]

What is the upper limit/range on the number of inflections a language might have in (poly)personal agreement, and what languages are examples?

For Yokai, I'm planning on polypersonal agreement. The language is also intended as quite fusional, such that the agreement for subject & object would be 1 affix. However, I also have 25 noun classes which means that there would be well over 600 suffixes just for personal agreement. Based on what I've seen, many Bantu languages have noun classes & agreement/concords for both subject & object, but the subject & object are separate affixes (or perhaps clitics?). That seems like the memory load would be much lower since there would be 10-20 affixes for subject & object each. They are suffixed separately since the languages are agglutinating.

To me it seems like 600+ forms for 1 suffix "slot" would be pretty unstable and simplified quickly, but maybe I'm wrong on that.

Any thoughts and/or suggestions?

Another option: use regular affixes, but add complex phonological rules on top of them, so the surface forms are effectively fused. This seems to be fairly common in many polysynthetic languages.

On the topic of direct-inverse marking, you may be interested in Oxford’s Algonquian Grammar Myths, which has several good sections at the beginning covering various parts of the Algonquian direct-inverse system. Most interesting to me is its general structure: one affix for the more animate actor, one affix for the less animate actor, and one direct-inverse affix. It occurs to me that such a structure would work well with Yokai noun classes: for instance, you could mark one actor with a macro prefix and the other with a micro suffix. This means you would need only two new affixes: one for direct, and one for inverse. (Or you could even get away with having direct be unmarked.)

[Vardelm]

Another option: use regular affixes, but add complex phonological rules on top of them, so the surface forms are effectively fused. This seems to be fairly common in many polysynthetic languages.

Possible, although I rather dread phonological rules. lol Maybe I shouldn't, especially since I plan to add consonant gradation to Dwarvish at some point. If these affixes are effectively fused, I will probably list them as though they are actually fused. That sort of leaves me in the original predicament, of course.

On the topic of direct-inverse marking, you may be interested in Oxford’s Algonquian Grammar Myths, which has several good sections at the beginning covering various parts of the Algonquian direct-inverse system.

I will take a look at that! I can always rely on you for quality references!

It occurs to me that such a structure would work well with Yokai noun classes: for instance, you could mark one actor with a macro prefix and the other with a micro suffix.

Huh. That's an interesting idea. I'll at least mull that over.

This means you would need only two new affixes: one for direct, and one for inverse. (Or you could even get away with having direct be unmarked.)

Plan is to have both marked since I want balance in both directions.

[bradrn]

Another option: use regular affixes, but add complex phonological rules on top of them, so the surface forms are effectively fused. This seems to be fairly common in many polysynthetic languages.

Possible, although I rather dread phonological rules. lol

Hah, yes, so do I! I’ve been trying to get over it by reading lots about phonological rules and sound changes.

On the topic of direct-inverse marking, you may be interested in Oxford’s Algonquian Grammar Myths, which has several good sections at the beginning covering various parts of the Algonquian direct-inverse system.

I will take a look at that! I can always rely on you for quality references!

You’re welcome! Though I’m sure I’ve actually linked this one before (in the ergativity thread, I think?).

[Ares Land]

For Yokai, I'm planning on polypersonal agreement. The language is also intended as quite fusional, such that the agreement for subject & object would be 1 affix. However, I also have 25 noun classes which means that there would be well over 600 suffixes just for personal agreement. Based on what I've seen, many Bantu languages have noun classes & agreement/concords for both subject & object, but the subject & object are separate affixes (or perhaps clitics?). That seems like the memory load would be much lower since there would be 10-20 affixes for subject & object each. They are suffixed separately since the languages are agglutinating.

Natlangs indeed don't help much in that much... I don't know of any that has both noun classes and fused subject/object markers.
Fulani (which is in the same ballbark as to the number of noun classes) has subject and object clitics, but these, unlike independant pronouns, don't mark noun classes at all.
Lingala just distinguishes animate/inanimate (and if I read the grammar right, doesn't mark the object anyway).

Bantu language that do have class concord on verb generally merge some of the noun classes for verbal concord.

To me it seems like 600+ forms for 1 suffix "slot" would be pretty unstable and simplified quickly, but maybe I'm wrong on that.

Any thoughts and/or suggestions?

An idea (though related to the animacy hierarchy and direct-inverse ones): the bulk of transitive sentences are going to involve a member of classes 1 to 5 as subject. And I'd bet most often classes 2 and 3.
What I'd do is have fused markers for the most common combinations (keeping in mind that noun classes can merge for verbal agreement). How many you keep is primarily a function of how much of these you can remember
To that, add the combined prefixes that end up identical through sound change.

Once you do that, there are a number of options for the less common combinations such as, indeed, direct-inverse.

[Vardelm]

Natlangs indeed don't help much in that much... I don't know of any that has both noun classes and fused subject/object markers.
Fulani (which is in the same ballbark as to the number of noun classes) has subject and object clitics, but these, unlike independant pronouns, don't mark noun classes at all.
Lingala just distinguishes animate/inanimate (and if I read the grammar right, doesn't mark the object anyway).

Yep. What I've seen is that they either mark both subject & object w/ independant, agglutinative affixes, or they mark the intransitive with the subject and the transitive with the object. For this lang, I don't want the subject & object to be able to be determined just from the agreement markers; I absolutely want the direct & inverse affixes (and animacy hierarchy) to drive that entirely.

Bantu language that do have class concord on verb generally merge some of the noun classes for verbal concord.

This is interesting. Do you have any particular ones in mind? I need to look into this since it would make me feel a little better about using the (plural) macro-classes since those would be a similar combination, although more regular.

An idea (though related to the animacy hierarchy and direct-inverse ones): the bulk of transitive sentences are going to involve a member of classes 1 to 5 as subject. And I'd bet most often classes 2 and 3.

Another good idea to mull over. This seems related to the merged concords you mention above.

[bradrn]

Natlangs indeed don't help much in that much... I don't know of any that has both noun classes and fused subject/object markers.
Fulani (which is in the same ballbark as to the number of noun classes) has subject and object clitics, but these, unlike independant pronouns, don't mark noun classes at all.
Lingala just distinguishes animate/inanimate (and if I read the grammar right, doesn't mark the object anyway).

Yep. What I've seen is that they either mark both subject & object w/ independant, agglutinative affixes … For this lang, I don't want the subject & object to be able to be determined just from the agreement markers; I absolutely want the direct & inverse affixes (and animacy hierarchy) to drive that entirely.

As far as I’m aware direct/inverse systems absolutely do mark both subject and object with independent affixes! The only difference is that, instead of one affix being the subject and one being the object, you have one affix being the more animate argument and the other being the less animate argument, so you need an extra marker to assign NPs to grammatical relations.

… or they mark the intransitive with the subject and the transitive with the object …

Are you talking about ergative verbal agreement here? Because otherwise I’m not quite sure what you’re trying to say here.

Bantu language that do have class concord on verb generally merge some of the noun classes for verbal concord.

This is interesting. Do you have any particular ones in mind? I need to look into this since it would make me feel a little better about using the (plural) macro-classes since those would be a similar combination, although more regular.

This sounded interesting, so I quickly looked at a Kikuyu grammar; it does show a bit of syncretism in the verbal prefixes, though not much (between classes 2/6, 3/14, 15/17). More interestingly, the verbal prefixes are often quite different to the nominal ones, especially in classes 1–5.

An idea (though related to the animacy hierarchy and direct-inverse ones): the bulk of transitive sentences are going to involve a member of classes 1 to 5 as subject. And I'd bet most often classes 2 and 3.

Another good idea to mull over. This seems related to the merged concords you mention above.

I think it has more to do with animacy: agents will predominantly come from the most animate noun classes. If anything, you probably don’t want to merge these noun classes with each other in the agentive argument affix, in order to make the finest distinctions amongst the most frequent cases. (Though you might want to merge them when used as undergoers!)

[Ares Land]

Bantu language that do have class concord on verb generally merge some of the noun classes for verbal concord.

This is interesting. Do you have any particular ones in mind? I need to look into this since it would make me feel a little better about using the (plural) macro-classes since those would be a similar combination, although more regular.

Swahili for instance: https://en.wikipedia.org/wiki/Swahili_g ... t_concords
Or Shimaore (closely related to Swahili within Bantu): http://ylangue.free.fr/chap_030.htm (in French but the table is clear enough).

Besides all animates tend to use class 1/2 affixes regardless of their actual noun class.

[Ares Land]

As far as I’m aware direct/inverse systems absolutely do mark both subject and object with independent affixes! The only difference is that, instead of one affix being the subject and one being the object, you have one affix being the more animate argument and the other being the less animate argument, so you need an extra marker to assign NPs to grammatical relations.

Ah, nope, I think that was about languages with polypersonal agreement and many noun classes

(There's no direct-inverse ,natural language with noun classes that I know of... though amusingly both Vardeim and I came up with the same idea for our conlangs.)

[bradrn]

As far as I’m aware direct/inverse systems absolutely do mark both subject and object with independent affixes! The only difference is that, instead of one affix being the subject and one being the object, you have one affix being the more animate argument and the other being the less animate argument, so you need an extra marker to assign NPs to grammatical relations.

Are you talking about ergative verbal agreement here? Because otherwise I’m not quite sure what you’re trying to say here.

Ah, nope, I think that was about languages with polypersonal agreement and numerous languages

Huh? What do you mean by ‘numerous languages’? (Also, I think that quote might have mangled my post a bit, given that each paragraph there was replying to a different part of Vardelm’s comments.)

(There's no direct-inverse ,natural language with noun classes that I know of... though amusingly both Vardeim and I came up with the same idea for our conlangs.)

Algonquian languages distinguish animacy, hence have noun classes. But yes, I know of no language with both a large noun class system and direct-inverse marking. (Both features are pretty uncommon, so that’s not surprising.)

[Ares Land]

Huh? What do you mean by ‘numerous languages’? (Also, I think that quote might have mangled my post a bit, given that each paragraph there was replying to a different part of Vardelm’s comments.)

Ah, sorry, I meant many noun classes (fixed my post)

[Vardelm]

As far as I’m aware direct/inverse systems absolutely do mark both subject and object with independent affixes! The only difference is that, instead of one affix being the subject and one being the object, you have one affix being the more animate argument and the other being the less animate argument, so you need an extra marker to assign NPs to grammatical relations.

You're correct about direct/inverse languages, and I'm quite familiar with how they work. What I'm referring to here are languages with noun classes (mostly African Bantu) that are not direct-inverse. The point is how they mark agreement with noun classes.

Are you talking about ergative verbal agreement here? Because otherwise I’m not quite sure what you’re trying to say here.

I wasn't specifically referring to ergative agreement, but I think you are probably correct. What I saw is that a transitive verb agrees with its subject while a transitive verb agrees with its object. Here, my point was that languages with noun classes don't seem to mark for both subject/agent and object with a single, fused affix like I am considering for Yokai.

This sounded interesting, so I quickly looked at a Kikuyu grammar; it does show a bit of syncretism in the verbal prefixes, though not much (between classes 2/6, 3/14, 15/17). More interestingly, the verbal prefixes are often quite different to the nominal ones, especially in classes 1–5.

I also find this interesting. When playing with different affixes to get a feel for how different mechanics might affect the aesthetic, I was trying to get the suffixes similar between noun & verb as that sort of felt more like what I expected from Bantu languages (although it's mostly prefixes there). That maybe doesn't need to be the case, which again suggests using the macro-classes for agreement could work. All depends on how it feels, though.

I think it has more to do with animacy: agents will predominantly come from the most animate noun classes. If anything, you probably don’t want to merge these noun classes with each other in the agentive argument affix, in order to make the finest distinctions amongst the most frequent cases. (Though you might want to merge them when used as undergoers!)

Yes, I would see it as an animacy issue as well, and I could see it going both ways. They could stay distinct because you're commonly using them to make the distinctions, or they get combined because they just phonologically slur together over time because they're used so often.

[Vardelm]

Ah, nope, I think that was about languages with polypersonal agreement and many noun classes

Correct, as I noted above.

(There's no direct-inverse ,natural language with noun classes that I know of... though amusingly both Vardeim and I came up with the same idea for our conlangs.)

Great minds.....

[bradrn]

As far as I’m aware direct/inverse systems absolutely do mark both subject and object with independent affixes! The only difference is that, instead of one affix being the subject and one being the object, you have one affix being the more animate argument and the other being the less animate argument, so you need an extra marker to assign NPs to grammatical relations.

You're correct about direct/inverse languages, and I'm quite familiar with how they work. What I'm referring to here are languages with noun classes (mostly African Bantu) that are not direct-inverse. The point is how they mark agreement with noun classes.

Ah, OK, thanks for clarifying!

Are you talking about ergative verbal agreement here? Because otherwise I’m not quite sure what you’re trying to say here.

I wasn't specifically referring to ergative agreement, but I think you are probably correct. What I saw is that a transitive verb agrees with its subject while a transitive verb agrees with its object. Here, my point was that languages with noun classes don't seem to mark for both subject/agent and object with a single, fused affix like I am considering for Yokai.

Interesting! This isn’t something I’d known before, but it sounds pretty normal based on what I know of other verbal categories. (For instance, verbal classifiers act the same way, as does pluractionality. It may have something to do with the fact that new arguments are usually placed in S or O.)

[Richard W]

Are you talking about ergative verbal agreement here? Because otherwise I’m not quite sure what you’re trying to say here.

I wasn't specifically referring to ergative agreement, but I think you are probably correct. What I saw is that a transitive verb agrees with its subject while a transitive verb agrees with its object. Here, my point was that languages with noun classes don't seem to mark for both subject/agent and object with a single, fused affix like I am considering for Yokai.

Did Romance flirt with ergativity? The perfect participles in the perfect tenses seem to have gone through a stage where transitive verbs had object-marking and habeo for the auxiliary verb and intransitive verbs has subject-marking and esse for the auxiliary verb..

[Vardelm]

Did Romance flirt with ergativity? The perfect participles in the perfect tenses seem to have gone through a stage where transitive verbs had object-marking and habeo for the auxiliary verb and intransitive verbs has subject-marking and esse for the auxiliary verb..

Maybe? No idea.

[Vardelm]

Deleting & reposting for visibility since this got pushed off the last page quickly.