Verb Complex Sketch (Requesting Feedback)

[leanancailin]

I'm working on a new language that has a larger and more involved verb complex that I've experimented with before, and would like some feedback and advice as I start moving into the details of how it all works and start up with the forms that the affixes actually take. I'm going to put up a brief sketch, then lay out the questions on my mind – though any input is welcome!

Evidentiality=Agreement(1)=Aspect(1)=ROOT=Aspect(3)=Aspect(2)=Tense=Direction=Voice=Causative=Applicative=Agreement(2)=Incorporation=Plural=Modal=Nonfinite

• Evidentiality is a binary direct/indirect system, with only indirects marked.
• Agreement is polypersonal – my original sketch has subject prefixes and object suffixes, but I'm learning towards making it a bit more complicated, with the subject prefixes changing shape for some objects / otherwise making object more of a transfix. In any rate, agreement is going to be 1/2/3 person only (with inclusive/exclusive) – no gender marking, and number marking comes later in the verb.
• Aspect is broken into 3 slots here, because I have several (mostly?) non-interacting aspect systems. The prefix is a simple perfective/imperfective (possibly with one unmarked). The suffix slot is broken up into aspects expressed with reduplication (full vs partial, expressing iterative and habitual) and those with suffix (a wider variety that, at the moment, includes continuative, inceptive, terminative, prospective, and durative). Both of the following types may be null.
• Tense includes future, present, and past with three levels of remoteness. Present (or perhaps recent past) may be unmarked.
• Direction indicates whether an action is moving towards, away from, or stationary relative to the object. This may change or expand in category (my inspiration languages also indicate things ”towards or from the sea” / ”towards or from shelter”).
• Voice can be unmarked for a typical active reading, or indicate a reflexive~reciprocal meaning and at least one passive – I'm playing around with including a separate voice with more of a middle meaning, an accidental separate from the passive, or a kind of resultative stative.
• I originally had causative and applicative included under voice, but I think they make more set as separate slots, to allow for things like a causative passive, or reflexive with an applicative. This may change back, or experience some sort of collapse or fusion.
• Incorporation is based on a (closed) system of object shape classifiers, which will be common with certain set expressions and ditransitive verbs, but otherwise mostly used to clarify an argument which isn't explicit in the clause.
• Plural marking indicates one (or more) of the arguments the verb agrees with is plural (or dual).
• Modal arguments (can, want, must) are most commonly indicates via suffixes here.
• The final nonfinite category includes an array of possible suffixes that transform the verb into a non-main clause, including a subjunctive morpheme, relativization, and several nominalizers.

Verbs also have pluractionality (indicated by a change in the root for a small, closed set) and all can be made into an intensive form (indicated by a form of preceding partial reduplication, which can be considered a part of the root for the purposes of this diagram)

My starting questions:

How does this system look as a template? Typologically, are there any elements that play poorly together, or any large implications for how the language functions?

Which of these aspects are likely to lean into each other and develop forms that are fused or exhibit high sandhi?

What are some examples of how real life languages handle internal slots that can have null expression? Wide use of allomorphs, use of brief markers incidcating 'nothing here', or just leaving it up to context and trusting the brain to puzzle it out?

How strictly do different languages assign canonical shapes to stems and affixes vs letting the shapes of words and syllables just play out? What kinds of strategies are common or reasonable to deal with roots that start or end with forms identical to common steam-adjacent prefixes or suffixes?

[Ares Land]

Caveat: I'm not an expert, at all. And I'm certainly not as competent as the Algonquian fans on this board but I've played with polysynthetic languages before.

About the two approaches I found that helped were:

• Try to figure out the diachronics of it, and how it could have came about through sound changes. Basically, you start with a simple template (possibly with a more analytic parent language) and try to figure out how elements could fit the structure. That way you get a plausible explanation for how your verb got that structure, plus you get all sort of interesting irregularities.
• Find out one or two good references grammar of a polysynthetic language and see how it handles any particular problem; this is often immensely eye-opening. (Nahuatl is fairly regular and not too complex, for instance, or on the opposite end of the spectrum, Iroquoian languages are insane monsters. Both are interesting.)

Really, a diachronic approach is best. Often I found that by deriving forms from an ancestral language and trying to make something regular out of the results I had stumbled upon a phenomenon common in existing languages but that I was unaware of.

How does this system look as a template? Typologically, are there any elements that play poorly together, or any large implications for how the language functions?

I don't really see any problem with it. I'd move incorporated roots closer to the verb myself, but if it makes sense diachronically, that's not a problem.

Which of these aspects are likely to lean into each other and develop forms that are fused or exhibit high sandhi?

Those that are most frequently used, I guess.
So agreement+aspect,aspect+tense, or voice+causative+applicative for instance.

What happens, basically, is that sandhi/fusion will merge any of your morphemes. This will lead to a high number of irregular patterns, the very common ones are kept, others fit together in a new regular template, and the least common ones will get regularized. (Think of English irregular verbs: they are or were those most commonly used.)

What are some examples of how real life languages handle internal slots that can have null expression? Wide use of allomorphs, use of brief markers incidcating 'nothing here', or just leaving it up to context and trusting the brain to puzzle it out?

French for instance mostly leaves it up to context, but allomorphy may help disambuate:
j'mange
j'ai mangé
j'ai pas mangé
tu manges
t'as manges
t'as pas mangé

What kinds of strategies are common or reasonable to deal with roots that start or end with forms identical to common steam-adjacent prefixes or suffixes?

I've no idea which strategies are most common, but stress related sound change or haplology seem reasonable, especially with common roots.

[bradrn]

About the two approaches I found that helped were:

• Try to figure out the diachronics of it, and how it could have came about through sound changes. Basically, you start with a simple template (possibly with a more analytic parent language) and try to figure out how elements could fit the structure. That way you get a plausible explanation for how your verb got that structure, plus you get all sort of interesting irregularities.

Do you think you could give a quick example of how you would do this? I quite like diachronics, but I don’t know much at all about how it interacts with syntax and morphology…

[akam chinjir]

I think it's pretty unusual to have valency-adjusting things (voice, causative, applicative) outside of TAM, though I don't know that it's unheard of. Some with incorporation.

(A lot of the time morpheme order corresponds to the semantics in an intuitive way. Like, for V+CAUS+PAST you get in the past (cause (V)); whereas you might expect V+PAST+CAUS to be cause (in the past (V))---tenselessly causing something to happen in the past, most likely not what you want.)

[Kuchigakatai]

How strictly do different languages assign canonical shapes to stems and affixes vs letting the shapes of words and syllables just play out?

Are you concerned about having too many irregular stems, or too many regularized ones?

I at least have been personally surprised by the amount of morphophonological change in Old French verbal stems, motivated by various things like stress movement (the stem vowel in [mɔˈvets ~ ˈmwɛvə(nt), paˈɾets ~ ˈpeɾə(nt)]), vowel loss leading to a consonant unacceptably in word-final position (devoicing in *[mwɛv] > [mwɛf]), vowel loss leading to an unacceptable cluster (dropping in *[mwɛfs] > [mwɛs], fortition in [ˈaʊdiːs] > *[ˈɔð(ə)s] > [ɔts]), retention of a pre-vocalic -j- through palatalization or metathesis ([ˈdɔlɛoː] > [ˈdwɔljo] > [dwɛʎ], [ˈpaːɾɛoː] > [ˈpaɾjo] > [paiɾ]).

mov(e)ō movēs *mov-ātis movent
> muef mue-s mov-ez muev-ent [mwɛf mwɛs mɔˈvets ˈmwɛvə(nt)]
lavō lavās lavātis lavant
> lef lev-es lav-ez lev-ent [lef ˈlevəs laˈvets ˈlevə(nt)]
amō amās amātis amant
> aim ain-s am-ez aim-ent [ãĩm ãĩns ãˈmets ˈãĩmə(nt)]
doleō dolēs *dol-ātis dolent
> dueil dueu-s dol-ez duel-ent [dwɛʎ dwɛws doˈlets ˈdwɛlə(nt)]
appellō appellās appellātis appellant
> apel apeau-s apel-ez apel-ent [aˈpɛl aˈpe̯aus apəˈlets aˈpɛlə(nt)]
teneō tenēs *ten-ātis tenent
> tieng tien-s ten-ez tien-ent [tjɛɲ tjɛns təˈnets ˈtjɛnə(nt)]
pāreō pārēs pārētis pārent
> pair per-s par-ez per-ent [paiɾ peɾs paˈɾets ˈpeɾə(nt)]
audiō audīs *aud-ātis aud(i)unt
> oi oz (i.e. ot-s!) o-ez o-ent [ɔi ɔts ɔˈets ˈɔə(nt)]
exeō exīs *ex-ātis ex(e)unt
> is is (i.e. is-s) eiss-iez iss-ent [is is eiˈsjets ˈisə(nt)]
parabolō parabolās par(abo)lātis parabolant
> parol paroles parlez parolent [paˈɾɔl paˈɾɔləs paɾˈlets paˈɾɔlə(nt)]

What kinds of strategies are common or reasonable to deal with roots that start or end with forms identical to common steam-adjacent prefixes or suffixes?

Spanish at least has not cared to do anything regarding lograr 'to achieve', lavar 'to wash' or separar 'to separate', even though their first syllable looks like a preverbal pronoun (lo, la and se, which are commonly used with them). So you get the likes of lo logré 'I achieved it, I managed to do it', la lavé 'I washed it (e.g. a shirt)', se separó 'it separated [from something]'.

Same goes for verbs that begin with syllables that look like prepositions, e.g. difícil de decir 'difficult to say', tardarse en enviarlo 'to delay sending it'.

About the two approaches I found that helped were:

• Try to figure out the diachronics of it, and how it could have came about through sound changes. Basically, you start with a simple template (possibly with a more analytic parent language) and try to figure out how elements could fit the structure. That way you get a plausible explanation for how your verb got that structure, plus you get all sort of interesting irregularities.

Do you think you could give a quick example of how you would do this? I quite like diachronics, but I don’t know much at all about how it interacts with syntax and morphology…

Something like Zompist's development of Hanying Creole (HC) into Modern Hanying (MH)?

Mariya nã tẽ meivi žedu kan [HC]
Mary not PAST maybe know.how.to read
'Mary may not have known how to read.'

In between HC and MH, the HC past-tense marker tẽ gets replaced by the completive emphatic sĩ > MH ze, and HC wala 'person, man' becomes the general 3rd person prefix lə-.

Mæri lənázmežeru kæn [MH]
Mæri lə-na-ze-me-žeru kæn
Mary 3-NEG-PAST-IRR-can read.INF
'Mary may not have been able to read.'

Etymologically, lənázmežeru is Hindi -वाला -vālā '(works-with-X-noun agent suffix)' + Portuguese não 'not' + Portuguese sim 'yes' + English maybe + Mandarin 知道 zhīdao 'know'.

[bradrn]

About the two approaches I found that helped were:

• Try to figure out the diachronics of it, and how it could have came about through sound changes. Basically, you start with a simple template (possibly with a more analytic parent language) and try to figure out how elements could fit the structure. That way you get a plausible explanation for how your verb got that structure, plus you get all sort of interesting irregularities.

Do you think you could give a quick example of how you would do this? I quite like diachronics, but I don’t know much at all about how it interacts with syntax and morphology…

Something like Zompist's development of Hanying Creole (HC) into Modern Hanying (MH)?

Mariya nã tẽ meivi žedu kan [HC]
Mary not PAST maybe know.how.to read
'Mary may not have known how to read.'

In between HC and MH, the HC past-tense marker tẽ gets replaced by the completive emphatic sĩ > MH ze, and HC wala 'person, man' becomes the general 3rd person prefix lə-.

Mæri lənázmežeru kæn [MH]
Mæri lə-na-ze-me-žeru kæn
Mary 3-NEG-PAST-IRR-can read.INF
'Mary may not have been able to read.'

Etymologically, lənázmežeru is Hindi -वाला -vālā '(works-with-X-noun agent suffix)' + Portuguese não 'not' + Portuguese sim 'yes' + English maybe + Mandarin 知道 zhīdao 'know'.

Good point, I’ll read through the Hanying grammar.

[Ares Land]

About the two approaches I found that helped were:

• Try to figure out the diachronics of it, and how it could have came about through sound changes. Basically, you start with a simple template (possibly with a more analytic parent language) and try to figure out how elements could fit the structure. That way you get a plausible explanation for how your verb got that structure, plus you get all sort of interesting irregularities.

Do you think you could give a quick example of how you would do this? I quite like diachronics, but I don’t know much at all about how it interacts with syntax and morphology…

Sure, let's say, for instance, that you have something like this in the proto-language:

niladat, I tell you / niladaki, we tell you

ni-lada-t

2s.OBJ-tell-1s

/

ni-lada-ki

2s.OBJ-tell-1p

Now, let's say final consonants are lost, and voiced consonants are lost intervocally while voiceless ones are lenited, and some non-stressed vowels are reduced:
So you get:
nlaa, tell you / nlaah / (PL) tell you
We lost subject agreement, but no worries: it's re-introduced with a cliticized subject pronoun:

wanlaa, I tell you, kitnlaah, we tell you
With further sound changes you get for instance:

wa:laa, I tell you, kinlaah, we tell you

And often enough your speakers say somethng like:

poda they say

po-da

tell-3p

Through sound change, it becomes:
poda > poa > po:

po:kinlaah

EVID-1p.2s-tell-PL

, 'We tell you (or so they say)' and so on.

You can start with an analytic language and agglutinate and fuse morphemes along the way. But it's more fun and more realistic, I believe to begin with a heavily inflected language, have it lose some inflections along the way and then reacquire them..., (Of course I'm influenced, in that respect, by French and Romance languages in general, which went through several cycles of this process.)

[bradrn]

About the two approaches I found that helped were:

• Try to figure out the diachronics of it, and how it could have came about through sound changes. Basically, you start with a simple template (possibly with a more analytic parent language) and try to figure out how elements could fit the structure. That way you get a plausible explanation for how your verb got that structure, plus you get all sort of interesting irregularities.

Do you think you could give a quick example of how you would do this? I quite like diachronics, but I don’t know much at all about how it interacts with syntax and morphology…

[elided for space]

I’m trying to derive another language from a heavily inflected protolanguage, so this is really helpful — thank you!

[Moose-tache]

How strictly do different languages assign canonical shapes to stems and affixes vs letting the shapes of words and syllables just play out? What kinds of strategies are common or reasonable to deal with roots that start or end with forms identical to common steam-adjacent prefixes or suffixes?

I'm not sure what the first part of this question means. Some languages with harmony will have over half a dozen versions of the same suffix based on harmonic class, if that's what you mean by "letting shapes just play out." And suppletion is pretty common cross-linguistically, so "cannonical shapes" may not even mean one thing.
As for the second part, how does English deal with this? The present tense of "to mind" is identical to the past tense of "to mine," but nobody seems to lose any sleep over "to mind" having part of its root show the same form as an inflectional ending.

[Richard W]

As for the second part, how does English deal with this? The present tense of "to mind" is identical to the past tense of "to mine," but nobody seems to lose any sleep over "to mind" having part of its root show the same form as an inflectional ending.

Mined can be longer than mind. Morpheme junctures aren't necessarily silent.

[Moose-tache]

I think you missed the point.

[Xwtek]

For the plural suffix, you can use adverb together or equally. (The latter is attested in Javanese)

Muride

student-DEF

padha

same

nyaut

ACT-answer

The students are answering (the teacher)

Also, Javanese apparently (according to a grammar book) has no plurality distinction in pronoun and in pronominal suffixes. (except for first-person)