Zompist Bboard Again

Undrivi is one of my old conlangs that I’m reworking, and I’d like your input on its vowel harmony system.

Undrivi has a two-parameter vowel harmony system where vowels harmonize according to both backness and closeness. It distinguishes eight vowel qualities, which are defined with regard to these parameters as follows:

i /i/: [+close], neutral backness
e /e/: [-close], neutral backness
ụ /ʉ/: [+close], [-back]
ọ /ɵ/: [-close], [-back]
u /u/: [+close], [+back]
o /o/: [-close], [+back]
ạ /a/: [-back], neutral closeness
a /ɑ/: [+back], neutral closeness

As you can see, while several vowels are neutral for one parameter, no vowel is neutral for both. The vowels thus form four harmonic sets:

i ụ ạ /i ʉ a/: [+close], [-back]
i u a /i u ɑ/: [+close], [+back]
e ọ ạ /e ɵ a/: [-close], [-back]
e o a /e o ɑ/: [-close], [+back]

We can therefore speak of three archiphonemes:

|I|: Harmonizes for closeness, realized as /i/ or /e/
|A|: Harmonizes for backness, realized as /a/ or /ɑ/
|U|: Harmonizes for closeness and backness, realized as /ʉ/, /u/, /ɵ/, or /o/

Any vowels in inflectional morphology, and most derivational morphology (a few morphemes break harmony) can thus be analyzed as one of these three archiphonemes.

As |U| is the only archiphoneme whose realization is different in all four harmonic environments, they can reasonably be named by referencing this vowel, thus there are ụ-, u-, ọ-, and o-type words. These need not necessarily contain the namesake vowel: dzafe, for example, is unambiguously an o-type word because it contains a [+back] and a [-close] vowel, and as such would take the o-type forms of inflections.

Any questions about how this works thus far? Does this system seem a) naturalistic, b) stable, and c) interesting?

Addendum:

In general the sets with greater acoustic difference between vowels are less marked, thus the largest number of words (~36%) are u-type, somewhat fewer words (~24% each) are ụ- and o-type, and the smallest number of words (~16%) are ọ-type.

Words containing only backness-neutral or only closeness-neutral vowels vary somewhat in what type of harmony their inflections take, but a majority of these words take the less marked option. Thus, roots with only /i/ usually take u-type inflections, roots in /e/ take o-type, roots in /a/ take ụ-type, and roots in /ɑ/ take u-type.

Chengjiang wrote: ↑Thu Aug 27, 2020 6:45 pm i ụ ạ /i ʉ a/: [+close], [-back]
i u a /i u ɑ/: [+close], [+back]
e ọ ạ /e ɵ a/: [-close], [-back]
e o a /e o ɑ/: [-close], [-back]

I think you wanted [-close] on that last one, correct?

Chengjiang wrote: ↑Thu Aug 27, 2020 6:45 pm Any questions about how this works thus far?

Are /i e/ transparent or opaque WRT harmony? That is, do additional vowels in before/after also harmonize, or do these vowels block any further advancement? If they are opaque, is the harmony progressive or regressive?

Chengjiang wrote: ↑Thu Aug 27, 2020 6:45 pm Does this system seem a) naturalistic

Probably? It seems like a good twist on a standard set of rules. I haven't seen anything quite like it, but it's just the combination of features that makes it different, so I don't see why it couldn't arise.

Chengjiang wrote: ↑Thu Aug 27, 2020 6:45 pm b) stable

I would think so. There's plenty of vowel space in there, so I don't see things shifting around too much compared to other systems.

Chengjiang wrote: ↑Thu Aug 27, 2020 6:45 pm c) interesting?

Definitely. I dig it.

Personally I would think about /a/ as being the vowel that shifts for backness & closeness rather than /u/, just because I feel like /a/ would be more common. That's a completely subjective evaluation, though. I don't see anything wrong with what you have.

Vardelm wrote: ↑Thu Aug 27, 2020 7:48 pm

Chengjiang wrote: ↑Thu Aug 27, 2020 6:45 pm i ụ ạ /i ʉ a/: [+close], [-back]
i u a /i u ɑ/: [+close], [+back]
e ọ ạ /e ɵ a/: [-close], [-back]
e o a /e o ɑ/: [-close], [-back]

I think you wanted [-close] on that last one, correct?

I did, and I’ll edit that. Thank you.

Are /i e/ transparent or opaque WRT harmony? That is, do additional vowels in before/after also harmonize, or do these vowels block any further advancement? If they are opaque, is the harmony progressive or regressive?

Neutral vowels of either type are transparent to harmony.

Personally I would think about /a/ as being the vowel that shifts for backness & closeness rather than /u/, just because I feel like /a/ would be more common. That's a completely subjective evaluation, though. I don't see anything wrong with what you have.

IA| is indeed more common than |U|, but not by a huge amount. And there’s precedent for /a/ being neutral despite being among the more common vowels in other harmony systems, e.g. many languages’ [ATR] harmony systems.

Thank you for your input!

I’m tempted to post about the language’s writing system, which is a sort of syllabary-abugida hybrid where the inherent vowel of a character (one of the three vowel archiphonemes) depends on the historical plain/palatalized/labialized status of the onset. It’s related to this and I think it’s interesting, but it’s obviously a rather in-depth topic in its own right.

Chengjiang wrote: ↑Thu Aug 27, 2020 7:59 pm IA| is indeed more common than |U|, but not by a huge amount. And there’s precedent for /a/ being neutral despite being among the more common vowels in other harmony systems, e.g. many languages’ [ATR] harmony systems.

Ah, sounds good. That seems to work really well for what you're doing, and obviously you've researched it a bit. Rock on!

Chengjiang wrote: ↑Thu Aug 27, 2020 7:59 pm Thank you for your input!

My pleasure.

Ofher tidbits about the vowel system:

All vowels can occur short and long. Long vowels are not distinct from sequences of identical short vowels, and can be analyzed as such, although historically a lot of them derive from sequences of a vowel plus a vocalized velar coda consonant, and the writing system represents them as the vowel plus this “chroneme” consonant. (It’s similar to the situation with <ğ> in Turkish.) Similarly, diphthongs are not phonemically distinct from heterogeneous vowel sequences, although most of them derive historically from sequences of vowel plus glide. (Most of them are versions of |AI| or |AU|; |IU| and |UI| diphthongs are relatively rare, and |IA| and |UA| collapse into |jA| and |wA| for many speakers.)

Undrivi has a phonemic pitch accent with two possible contours, rising and falling. On short vowels that are in open syllables or have an obstruent coda, the rising pitch is realized as a low tone on the accented vowel with a high tone on the next syllable, while the falling pitch is realized as a high tone on the accented vowel with a low tone on the following syllable. On long vowels or on short vowels with sonorant codas, the rising pitch is realized as a low-to-high rising tone, while the falling pitch is realized as a high-to-low falling tone. The unaccented syllables have more mid tones, approaching the tone of the accented syllable or post-accented syllable as they get nearer to it. Thus, with similar nonsense syllables for comparison, where 1 is the lowest pitch and 5 is the highest:

akáta — 2-1-5
akàta — 4-5-1
akásta — 2-1-5
akàsta — 4-5-1
akáata — 2-15-4
akàata — 4-51-2
akánta — 2-15-4
akànta — 4-51-2

Unaccented words do exist, and they include both particles and content words, although it’s rare for words of four or more syllables not to have an accent. Short unaccented words tend to have a pitch contour determined by neighboring words, while longer ones (mostly certain inflectional forms) tend to have an overall downward contour without any steep changes in pitch.

I’m particularly concerned about stability since

1) I’m aware of two-dimensional harmony systems, but only where one of the dimensions is roundness, e.g. Turkish. I don’t know of any natlangs that quite have this system.

2) Both closeness and backness harmony somewhat resemble ATR harmony, and so I could see this system decaying into a one-dimensional ATR harmony system.

This looks like some kind of cross-breed between Mongolian and Turkish. I think you should show examples of what happens with different suffixes when they are added to different stems. For example, if a stem only contains /i/, what does a suffix with |A| turn into?

Chengjiang wrote: ↑Thu Aug 27, 2020 6:45 pm Any questions about how this works thus far? Does this system seem a) naturalistic, b) stable, and c) interesting?

I think this looks naturalistic. It is worth mentioning that some Bantu languages with ATR harmony also have a few non-productive cases of backness harmony, IIRC. I also think it looks stable and definitely interesting, since not many languages combine ATR harmony with backness harmony.

I have a question on your archiphonemes and an alternative formulation of your vowel harmony. For example, could you also say that /i/ and /e/ are neutralized in inflectional morphology, since their closeness value is overwritten by the root vowel? This would be an alternative to the archiphoneme |I|. You could explain the non-application of backness harmony here by the lack of a [+back] counterpart of these vowels. This reformulation would then also work for |A| and |U|. Not that you need to change the description. This is just for me to better understand the pattern.

Thanks for the feedback! It’s clear some examples are in order. For reference, I’ve included the language’s consonant inventory at the bottom of the post, although it’s not necessary for understanding these examples.

The example words are gwirú “toad”, dzafe “sign”, kạvli “bowl”, dọldé “leaf”, mìs “glass”, téf “fire”, fạlạạ “cliff”, and tsakán “pebble”. The inflections we’ll use to show vowel harmony are the plural ending -(I)nA and the genitive -(y)U, as well as their combination -(I)nAU. (Yes, Undrivi is primarily agglutinating, with a low degree of fusion.) Thus:

gwirú, gwirúyu, gwirúna, gwirúnau
dzafe, dzafeyo, dzafena, dzafenao
kạvli, kạvliyụ, kạvlinạ, kạvlinạụ
dọldé, dọldéyọ, dọldénạ, dọldénạọ
mìs, mìzu, mìzna, mìznau
téf, téfo, tévna, tévnao
fạlạạ fạlạụ fạlạạnạ fạlạạnạụ
tsakán tsakántu tsakántina tsakántinau

As you can see, these words include each of the four harmonic possibilities and show what (most often) happens in words with only one of the neutral vowels. You may also note some stem changes, although they aren’t really relevant to this topic.

Creyeditor, that is indeed another way you could analyze the vowels. I prefer to talk of archiphonemes as I find it
makes it easier to discuss the base form of inflections.

——

Undrivi’s consonant inventory:

m n nh /m n ɲ/
p b t d k g kw gw /p b t d k ɡ kʷ ɡʷ/
ts dz ch j /ts dz tʃ dʒ/
f v s z sh zh /f v s z ʃ ʒ/
r /ɾ/
l lh /l ʎ/
y w /j w/

There are also a series of consonants nw tw dw sw zw that were historically labialized alveolars but have become something different in every major dialect. A full look into what they are, along with a full discussion of the language’s allophonic processes and phonotactics, is probably best left for another thread. For now, I’ll just note that the language allows at most CVC syllables with a restricted set of finals and none of these consonants are realized dramatically differently from this in the words I’ve given.