Zompist Bboard Again

I'm finding that there are too many diachronic and morphological issues that make me uncertain whether my language is really triconsonantal or just has a really complex ablaut system.

For starters, I was reading this old thread about creating a triconsonantal root language and given what it said about the vowels, I'm concerned that my pre-triconlang's vowel system might be too large, as it has: Which is at least four vowels more than the reccommendation. However, I need it to be that way in order to maximize the phonemicization of my palatal series.

Mostly, it's my patterns/templates that make me unsure that my conlang is really triconsonantal, as really only one vowel gets changed rather than two or more and I don't know if it's my pre-tricon morphology that is hindering the development: the end result of these things after some diachronic processes is supposed to be something like this (I should note that long vowels do not get elided by syncope):
Verb patterns:
C₁VC₂uC₃ (active voice singular)
C₁VC₂C₃əm (active voice plural)

C₁VC₂uC₃ (passive voice singular)
C₁VC₂C₃əm (passive voice plural)

tarC₁VC₂uC₃ (applicative-active voice singular)
tarC₁VC₂C₃əm (applicative-active voice plural)

tarC₁VC₂uC₃ (applicative-passive voice singular)
tarC₁VC₂C₃əm (applicative-passive voice plural)

Adjectives/Adverbs:
C₁əC₂C₂aC₃-
C₁əC₂C₂uC₃-
C₁əC₂C₂iC₃-
C₁əC₂C₂əC₃-

Nouns:
https://conworkshop.com/view_article.ph ... 749f1b4cc3
Also:
https://conworkshop.com/view_article.ph ... 529b7aa32d

I don't know... it just, doesn't seem "triconsonatal" enough. There just isn't enough ablaut or analogizing of forms it seems.

One thing I'll ask you right off the bat: how many roots have you put through this set of transformations? The more information available, the easier you may find it to locate things to tweak.
(Also: the dividing line between a complex ablaut system and a triconsonantal root language is relatively fine. Just keep it up, maybe add a few more changes if you feel it necessary. Also, you can always try coming to the vowels you desire through the simplification of diphthongs--you might find it opens up a few more opportunities for change...)

Can't help with the main goal, here... Semitic may have gotten where it is very slowly, with little baby steps...you might be on the right path, but its a long path for sure.

The aspect names are interesting.

Pedant wrote: ↑Tue Jul 02, 2019 9:49 pm One thing I'll ask you right off the bat: how many roots have you put through this set of transformations? The more information available, the easier you may find it to locate things to tweak.

Not a whole lot, I do a lot of back-and-forth between the current lang and the pre-tricon ancestor. There's too many triliteral roots I can't account for diachronically.

Also, you can always try coming to the vowels you desire through the simplification of diphthongs--you might find it opens up a few more opportunities for change...)

The pre-triconlang is strictly CV(:)(C), so I can't really do vowel clustering.

It took me years and years to get a tricon language I was satisfied with. Two things that really helped me: 1) reading Christopher Ehret's Reconstructing Proto-Afroasiatic (Proto-Afrasian)--his ideas are controversial, but just because the real Semitic didn't develop that way doesn't mean his ideas can't be used in a constructed language--and 2) learning how real Semitic languages work (Phoenician/Punic in my case). One major point is that Semitic heavily levels sound changes through analogy, especially in verbs; don't be afraid to reject what SCA gives you in favor of matching the pattern to your other verbs. Semitic languages can have irregular verbs (weak verbs in Northwest Semitic spring to mind), but even they follow a particular pattern.

well I guess take a look at the Wikipedia's article for Biblical Hebrew, modern Hebrew Grammar and Arabic grammar, including introductions to the broken plural in Arabic.

btw, I recently have also started a semitic-like language with triliteral roots.

Your first step might be expunging the expression "triconsonantal root language" from your thoughts entirely, the same way we've mostly managed to stop all talking about making "a polysynthetic language".

Aim for your language to be a certain way, not for it to be described by certain labels. Because the labels are arbitrary and meaningless. "A triconsonantal language" just means a language with a lot of ablaut, few phonemic vowels, and a tendency to have roots with three consonants on average. It's nothing magical. There is no dividing line that rings a bell and flashes a light when, yes, you have achieved a triconsonantal language and, oh no, sounds a sad-bassoon buzzer when you haven't.

It's legitimate and sensible, of course, to want to emulate Semitic specifically. But I think you already know that your languge's system is not as extensively "tricon" as Semitic.

If you're not trying to make specifically a Semiticoid clone, then there is no way to tell if your language is "triconsonantal enough", and it's a question that will only drive you mad with second-guessing yourself.

Do YOU want there to be more ablaut? Then add more ablaut. Do you want there to be less ablaut? Then subtract some ablaut. Asking us whether WE think it is triconsonantal "enough" will not yield any meaningful answers!

Salmoneus wrote: ↑Wed Jul 03, 2019 1:59 pm Your first step might be expunging the expression "triconsonantal root language" from your thoughts entirely, the same way we've mostly managed to stop all talking about making "a polysynthetic language".

Why would I expunge a useful term/concept? Anybody who hears me say it understands what I'm referring to. And it's very clearly distinct from what proto-Indo-European does (mostly because it took it farther).

I want more ablaut but I don't know how to add more. The only things that create ablaut effects so far are the active and passive voice. I also want vowel mutations to extend beyond just one vowel. At the moment my verb system is more or less just a polysyllablic version of "run vs. ran" rather than "katab vs. kaatib vs. katiib vs. kutub". Only one vowel is being mutated or alternated. Not more. Triconsonantal languages mutate more than just one vowel at a time, likely by analogical and morphological processes that I simply cannot imagine or figure out. And that's my major problem.

The goal is that root vowels of words are supposed to be eroded and analogized away from existence and I don't think my current set up is allowing that. And THAT is what I mean by "triconsonantal enough". There's people here who understand tricon languages and analogy and mutation better than I do, and are thus better able to identify such structures, so asking if they know if it's tricon enough would be fruitful.

I really can't fathom how you don't know how to add more ablaut. You just.... put in more ablaut. If you want your language to be a certain way, you do that. What's so hard about it? Is it justifying it diachronically?

One of my conlangs, Laqar, is not a triconsonantal language and does not try to be one, but achieves a similar effect through combining mobile stress based upon syllable weight, which is subsequently obscured by loss of phonemic vowel length, with systematic elision of unstressed vowels in certain positions, conditional reduction or vowel-coloring of vowel-semivowel sequences, and conditional elision of consonants to obscure many of the other changes, and in such a fashion that all the sound changes are very plausible, and none of them, whether by themselves, come off as unnatural or contrived. I would not try to just force a language into a triconsonantal model but rather try to build a language based upon things like what I used to build Laqar that together would result in considerable synthesis and which would result in significant surface "irregularity" even when all the sound changes behind it are entirely regular.

Ahzoh wrote: ↑Wed Jul 03, 2019 3:47 pm

Salmoneus wrote: ↑Wed Jul 03, 2019 1:59 pm Your first step might be expunging the expression "triconsonantal root language" from your thoughts entirely, the same way we've mostly managed to stop all talking about making "a polysynthetic language".

Why would I expunge a useful term/concept?

Because it's not a useful term. "Semitic-like" would be a useful term, but "triconsonantal" really isn't.

Anybody who hears me say it understands what I'm referring to.

But the five thousand topics we've had over the years of people asking "is this triconsonantal??" demonstrate that, no, they don't. And since you know all the facts about your language, if you're still asking if it's really a triconsonantal language by definition means that YOU don't know exactly what you mean by it either!

And it's very clearly distinct from what proto-Indo-European does (mostly because it took it farther).

But taking things "farther" isn't something that lends itself to objective labels. If A is only farther along than B, then you're talking about a continuum, not discrete categories, which makes the question of definition arbitrary.

The goal is that root vowels of words are supposed to be eroded and analogized away from existence... And THAT is what I mean by "triconsonantal enough"

OK! That's something concrete we can work with (albeit something rather depressing). That's something we give a concrete answer to. Except...

and I don't think my current set up is allowing that.

...you already know the answer!

Have the root vowels of words in your language been eroded and analogised away from existence? No, clearly they haven't, as you say repeatedly. And you say explicitly that you recognise that they're not and can't be under your "current set up".

So the question you think you're asking is one you've already answered for yourself, definitively and unambiguously. If your goal is to have a language in which vowels have no phonemic weight in distinguishing roots, then we all know that you have not met that goal*. So what answer are you looking for?

*indeed, at present you seem to have LESS ablaut than, say, Germanic has. Although admittedly I don't understand your description.

I want more ablaut but I don't know how to add more. The only things that create ablaut effects so far are the active and passive voice. I also want vowel mutations to extend beyond just one vowel. At the moment my verb system is more or less just a polysyllablic version of "run vs. ran" rather than "katab vs. katiib vs. kutub". Only one vowel is being mutated or alternated. Not more. Triconsonantal languages mutate more than just one vowel at a time, likely by analogical and morphological processes that I simply cannot imagine or figure out. And that's my major problem.

Now THIS is a much more fruitful line of inquiry, but it's got nothing to do with the question you asked. So instead of asking something vague and unanswerable like "is this a triconsonantal language enough?", ask something concrete and specific like "how can a language reduce the phonemic load of root vowels?"


A good start would be reading the post on the old board that you linked to. [You may also want to check out Mak's old thread in the L&L Museum on how semitic languages actually work] It pretty much answers your questions.

-----------------

But to put it in a few bulletpoints

- your goal is to reduce the number of phonemic vowels in roots to zero, or at most one. So obviously it helps to begin with as few vowels as possible.

- if you have too many vowels, you could reduce them by moving some properties to adjacent consonants. For example, you can easily reduce vowel inventories by introducing palatalisation and/or labialisation.

- you then need to have the realisation of the vowel in the root be influenced by affixes and/or stress movement. The most obvious ways to do this are vowel harmony and reduction. Although there are also other ways (eg harmonic effects triggered by consonants).

- this establishes ablaut patterns. You then just need to reduce the number of ablaut patterns to one, by having rarer patterns analogise (and/or by continuing to reduce the vowel inventory per se).

- to draw attention to the nonconcatenative morphology, you may then want to erode and merge some of the triggering affixes.


--------


For example, let's say we have four vowels, /a e i u/, and most words have the shape CVCVC. Important affixes for verbs include:

-aq (past tense)
-i(j) (imperfective aspect)
-in (agent derivative)
-ak (patient derivative)
-uns (locative derivative)
-et (plural subject agreement, also plural nominal marker)


We might (and this is only an example!) then....

a) before even adding the suffixes, reduce the number of vowel combinations through harmony. Before /u/: /a/ > /o/; /e/ and /i/ merge as /y/. Before /i/: /e/ > /i/, /a/ > /e/, /u/ > /y/. Before /a/: /e/ > /a/, /u/ > /o/. This leaves only one front-back pairing and no /y/ in final syllables, so some progressive, allophonic harmony: /u/ > /y/ after /y/.

b) put stress on initial syllable, reduce unstressed vowels slightly (y>i, e>a). This gives us the following pairings: a-a, e-i, i-a, i-i, y-i, o-a, o-u, u-u.

c) raise /o/ to /u/, which only merges two pairings. Also, /y/ and /u/ are no longer phonemically contrastive, so front all /u/ to /y/.

d) labialise consonants adjacent to /y/ (the only rounded vowel), and merge /y/ with /i/. This gives you the pairings a-a, e-i, i-a, i-i. Specifically, you'll note that of the 16 possible initial vowel pairings, 13 of them now have /i/ as the second vowel and only three have /a/ as the second vowel. So, the few remaining exceptions analogise to have /i/ also.

e) stress moves to the final syllable. Pre-stress vowels reduce to schwa.

f) now add the suffixes.

g) umlaut (throughout word): before /e/ or /i/: /a/ > /e/, /u/ > /i/. Before /a/: /i/ > /e/, /u/ > /o/. Before /u/: /i/ > /u/.

h) stress moves to penultimate mora; secondary stress projected two morae earlier.

i) as you can see, of the initial 16 pairings, 14 have merged into a single vowel pattern, so the remaining two do likewise.

j) unstressed /a/ raises to /e/

k) unstressed non-final vowels reduce to schwa; unstressed schwa drops outside of bisyllabic words, except where this would create a cluster within a syllable, and remaining schwa merges with /e/

l) tongue root harmony triggered by uvulars (which merge with velars), followed by vowel mergers to yield /i/ > /e/, /e/ and /o/ > /a/, /u/ > /o/.

n) loss of coda /n/

m) assorted consonant changes, including /s_w/ > /S/

o) Vje > V:

Thus, if we an example word, we have:

supak > S-P-K, "to sleep":
spik - he sleeps
spakak - he slept
spiki - he is sleeping
spekeki - he was sleeping
sepki - sleeper
spekek - period of sleep (that that is slept)
sepkush - bed (kw>k by analogy)
spiket - they sleep
sapkakat - they slept
sepki:t - they are sleeping
spakfatat - they were sleeping (kk>kx>kf)
sepkinet - sleepers
sepkeket - periods of sleep
sepkishet - beds

And the same pattern obtains for all other triconsonantal roots. And so forth.


[obviously I'm not saying this is either the specifi root Semitic took, nor the simplest root you could take, just an intereting one that was on the top of my head]

KathTheDragon wrote: ↑Wed Jul 03, 2019 4:36 pm I really can't fathom how you don't know how to add more ablaut. You just.... put in more ablaut. If you want your language to be a certain way, you do that. What's so hard about it? Is it justifying it diachronically?

I'm not skilled with diachronic like you might be. "Just put ablaut" is not so simple to me.

And yes, partly I don't know how to justify some mutations and analogies diachronically. I want to do the kind of things that Arabic does with their broken plurals, only between nouns in the governed state and nouns in the construct state, but there is no information that tells me how those patterns arose in that way.

My verb conjugation paradigms and noun case and gender system is different so I won't be following the exact path of Semitic languages, but that's ok. That's not my issue.

Salmoneus wrote: ↑Wed Jul 03, 2019 5:36 pm Have the root vowels of words in your language been eroded and analogised away from existence? No, clearly they haven't, as you say repeatedly. And you say explicitly that you recognise that they're not and can't be under your "current set up".

You misread. I said "I think" which confers the meaning of "I'm not sure" to the phrase. I do not actually know the answer because I'm not a diachronics and morphological genius like you lot are.

Now THIS is a much more fruitful line of inquiry, but it's got nothing to do with the question you asked. So instead of asking something vague and unanswerable like "is this a triconsonantal language enough?", ask something concrete and specific like "how can a language reduce the phonemic load of root vowels?"

I start off with a general question so that I can eventually ask more specific questions as elaboration is required. You may find that unhelpful, but I'm not good at articulating my issues.

I don't really want the vowel qualities and lengths to change much, I want the descendants to do that later.

A good start would be reading the post on the old board that you linked to. [You may also want to check out Mak's old thread in the L&L Museum on how semitic languages actually work] It pretty much answers your questions.

I have read those threads and the one I linked to and even read the Unfolding of Language and none of those help me much. Those things are really only helpful if I have the same nominal and verbal morphology as Semitic.

I don't know, for instance, how to come up with turning long words like "panedná" and "pindārā́zen" to "paned" and "pindās" like Tiramisu is so effortless able to do.

Honestly? My sort-of-consonantal-root-language isn't even diachronic-first. I make the patterns I want first and then bullshit out justifications after the fact, because I know from experience that doing it the other way round is headache-inducing. I haven't even gone all the way back to pre-consonantal-root times, because that's also headache-inducing. I just want a language, y'know?

KathTheDragon wrote: ↑Wed Jul 03, 2019 6:17 pm Honestly? My sort-of-consonantal-root-language isn't even diachronic-first. I make the patterns I want first and then bullshit out justifications after the fact, because I know from experience that doing it the other way round is headache-inducing. I haven't even gone all the way back to pre-consonantal-root times, because that's also headache-inducing. I just want a language, y'know?

Kind of what I'm doing as well. But I also like to have some reason for why a noun in the governed state is ilana but the construct state for it is ulun. I can at least diachronically justify why sewwa and reyye become sip and ruk in the construct state.

There's only so much handwaving I can do before it doesn't feel right.

On the other hand, diachronicalling it like tiramisu did can create easy complex patterns if I knew how to shorten long words to that degree.

KathTheDragon wrote: ↑Wed Jul 03, 2019 6:17 pm Honestly? My sort-of-consonantal-root-language isn't even diachronic-first. I make the patterns I want first and then bullshit out justifications after the fact, because I know from experience that doing it the other way round is headache-inducing. I haven't even gone all the way back to pre-consonantal-root times, because that's also headache-inducing. I just want a language, y'know?

Yeah, something I've had to accept is that I can't diachronically explain every morphosyntactic pattern in a conlang - unless I want to start from a very simple, pidgin-like creole (but even then...) - which is something I've done before, but it's silly to do that every time, because there are plenty of real-life languages whose morphology cannot be traced back to the point of grammaticalization. Proto-Indo-European's immediate ancestor wasn't a creole; it has morphological structures that likely date back indefinitely into the mists of time. Many languages have morphological structures that are so stable, so old, so thoroughly maintained and constantly readjusted through analogy that their ultimate origin cannot be traced; they may as well jave existed forever. The ablaut patterns of Semitic are probably like this to an extent.

So for OP, when it comes down to it, you can just put in whatever structures you want without tracing its diachronic origin - after all, you'd then have to trace the origin of the origin. Something has to just be arbitrary at some point.

missals wrote: ↑Wed Jul 03, 2019 9:28 pmYeah, something I've had to accept is that I can't diachronically explain every morphosyntactic pattern in a conlang

it can be done .... takes a lot of time, but it can be done. What starts out simple can become complex, too.

Im not trying to weigh you down. But I do have to be honest .... i started Pabappa with a randomly generated series of affixes for all its inflections and said I'd work out the diachronics later. But I just wasnt happy that way, and none of what I saw really made sense. I was trying too hard to be random, when in reality, what looks random superficially usually has a hidden pattern. so i had to go start everything all over, and derive the inflections from ~7000 yrs of grammar and sound changes.

Some of this may depend on the type of language ... with Pabappa I couldnt just fake it and say "oh, the -ā got analogized to -o" because there were very few analogical changes like that in the main vocabulary..... to have massive analogy in the grammar would be suspicious. But in Moonshine, although Im also doing it the hard way, I admit that my final product may not look different than what I had started Pabappa with.

I use a lot of throwaway morphemes, though. For example, the person markers in Poswa (the language in my sig) are derived from *evidentials*, of all things. the parent language marked person on verbs with -ap/-as/-a for 1st/2nd/3rd person. (i'm ignoring plurals for now). But it had evidentials -bu/-bi/-ba for "feel"/"see"/"know". and these evolved into the new person markers, which through regular sound change became -o/-e/-a.

On a wholly separate strain of thought, you could derive biconsonantalism from CVC+V ---> CVVC (metathesis) ---> CV²C, and then have separate morphemes that add both a third consonant and a second vowel. This might be more true to the Semitic model, anyway, since Ive heard it proposed that Semitic was originally biconsonantal and that some biconsonantal roots remain even today.

for example suppose /kat/ = water, and -i = locative case.
then /kati/ "in the water" > /kait/ > /kēt/. maybe later there is ē>i and a new extra syllable to disambiguate the resulting collisions, since by this point the original vowel will have become completely obscured. the thing about this is that having a large vowel inventory to begin with might actually be a positive thing, since it's almost certain that you'll have to compress it as you form the diphpthongs.

Icecap Moonshine has gradations like /ňap/ "speak" --> /nef-ř/, where the whole stem changes just from adding a suffix. this isnt tricon but a similar idea perhaps, since the ř could just disappear. this wouldnt be tricon, but still would be nonconcatenative, which is similar. yet the ňap>nef arose from palatalization that bled out in both directions.

you might find https://en.wikipedia.org/wiki/Yurok_language#Morphology interesting too, not that i or anyone else here can tell you where the patterns come from ... its anyones guess.

The California languages you should be looking at for inspiration for a nonconcatenative system are really Yokutsan languages, which have very, very extensive nonconcatenative morphology. (Just be careful that you only use sources based on actual personal fieldwork; all the literature on the languages as case studies in an OT or nonlinear framework or whatever using second-hand sources is useless because it's based on imaginarily constructing like 95%* of the "examples" based on rules formulated by Stanley Newman in the 1940s for [primarily] Yowlumni Yokuts.)




*This is an exaggeration, but not by much. William Weigel in "Real data, contrived data, and the Yokuts Canon" found that in the sources he tested, "The weighted average [contrivance rate] percentage was 76.2 percent of the total of all forms [cited]" and "No article had a contrivance rate lower than 65 percent." (And there are further problems with the way the contrived data is used that Weigel discusses.)

I should note that I’m already familiar with all the different types of nonconcatenative morphology, like reduplication, vowel lengthening, consonant gradation, etc.

Whimemsz wrote: ↑Fri Jul 05, 2019 1:56 pm The California languages you should be looking at for inspiration for a nonconcatenative system are really Yokutsan languages, which have very, very extensive nonconcatenative morphology. (Just be careful that you only use sources based on actual personal fieldwork; all the literature on the languages as case studies in an OT or nonlinear framework or whatever using second-hand sources is useless because it's based on imaginarily constructing like 95%* of the "examples" based on rules formulated by Stanley Newman in the 1940s for [primarily] Yowlumni Yokuts.)




*This is an exaggeration, but not by much. William Weigel in "Real data, contrived data, and the Yokuts Canon" found that in the sources he tested, "The weighted average [contrivance rate] percentage was 76.2 percent of the total of all forms [cited]" and "No article had a contrivance rate lower than 65 percent." (And there are further problems with the way the contrived data is used that Weigel discusses.)

Do you have any primary sources?