Random phonology/phonemic inventory thread

[DesEsseintes]

Can’t sleep so decided to post this.

This is inspired by - but hopefully differs substantially from - the wildly popular Akiatu over on the ZBB:

/m n ŋ/ m n ng
/p t k q ʔ/ p t k q ’
/f s x/ f s h
/v r ɣ/ v r g
/j w/ j w

/a e i ɔ u ʌ ɯ/ a e i o u ė į

e only occurs in the combinations je we, or after a coronal (n t s r).

There are probably some diphthongs, including ai au.

Syllabic structure favours open syllables with very few clusters beyond CG. The only permissible word-internal coda (for now at least) is the glottal stop.

Permissible consonant+glide sequences are as follows*:

Code: Select all

mj  mw          ngj ngw
pj  pw          kj  kw  qj  qw  ’j  ’w
fj  fw          hj  hw
vj  vw  rj  rw  gj  gw

*This distribution is probably the greatest Akiatu influence on - and inspiration for - the phonology.

[Birdlang]

/p b t d k g/ p b t d k g
/f v s z ç ʝ x ɣ ʜ ʢ h/ f v s z š ž x ǥ ḥ ḩ h
/m n ɳʲ ɳᵍ/ m n ñ ŋ
/ð j ɰ~w/ ð j w
/r/ ř
/ɾ/ r
/l/ l
/ɮˁ~ðˁ/ ẕ

/i y u ʊ/ i y u û
/ʉ ɜ ɒ̞̈/ ů ĕ ä
/e o ɛ ɔ a/ e o ê ô a
/ˈaː ˌaː aː ˈa ˌa/ ā́ ā̀ ā á à

[Frislander]

/p t t͡s t͡ʃ k/
/s ʃ/
/m n/
/ʋ l j/

/i iː u uː/
/eː oː/
/ɛ æː ɐ ɑː/

Syllable structure is (C)(R)V(R)C, where initial clusters are restricted to /p t t͡s t͡ʃ k s ʃ m n l/ + /j/, /t t͡s t͡ʃ k s ʃ/ + /ʋ/ and /p t k/ + /l/, and final consonants are restricted to /s ʃ m n l/, or the clusters /p k m n l/ + /s ʃ/.

[Auvon]

dgriddlang, australia-style obv

feet are bisyllabic with onset to σ mandatory, so [CV((C)V)] is a foot.

/p t̪ t̺ ʈ c k/ full grade
/β ð ɹ̺ ɻ j ɣ/ soft grade - can't appear root-initially except via gradation (eg: |ügl-kú| has the prefix become an infix for phonotactic consraints, so /güglú/(not ipa to the left, ipa to the right) [ɣɨ̀ʎó], with soft gradation triggered by /gl/ on |k|)
/m n̪ n̺ ɳ ɲ ŋ/ nasal grade
/l̪ l̺ ɭ ʎ ʟ/ laterals - do not participate in gradation
/i ɨ u ə a ɔ̃ i: ɨ: u: ə: a: ɔ̃:/ word level height harmony between a and ä. some lects merge o into this pair of harmonizing vowels
/á a/ tone (well, marked tone; everything has a surface tone ofc, which is low if unmarked) can only appear zeronce or once per foot. ocp prefers that two marked tones are not adjacent, but this is not a highly ranked constraint

paths of gradation are full<>soft(>zero, which is just soft C between two short V becomes VV if this does not violate mandatory foot onset), and full<>nasal. laterals can interact with but not participate in gradation, as seen above

ortho:
<p th t rt tg k>
<b zh z rz j/y g>
<m nh n rn gn ng>
<lh l rl gl lg>
<i ü u ä a o i: ü: u: ä: a: o:> alternatively <e> instead of <ä>
<á a>

[Shemtov]

Mos, an isolate spoken in North Iran. Based on the name, folk belief is that al-Khiḍr was from their ancestors, and the name was given to them by interaction with the Banu Israil, given al-Khiḍr's association with Musa:
/p b ɓ t d ɗ t͡ʃ d͡ʒ k g q/
/m n/
/ θ ð s z ʃ ʒ x ɣ h/
/r/
/l j ʋ/

/i u e o a/

Arabic writing (reversed):
​​​​​<پ ب بّ ت د دّ چ ج ک گ ق>
<م ن>
<ث ذ س ز ش ژ ح ع ه>
<ر>
<ل ی وْ>

<ي و اِ اُ ا>

[DV82LECM]

"Mos" in Crona means BAD. But this is GOOD.

[Porphyrogenitos]

So, a concept based on some things I learned about in some recent classes. A while ago I read about tone sandhi, and how in many Sinitic tone sandhi systems, you'll have a thing where every syllable in a syntactic phrase takes the sandhi-tone except for the last, and how each tone essentially has one sandhi form that it switches to when in a sandhi position.

The other day a professor of mine was talking about French stress and how it's phrase-level, with no stress at all except the last syllable in a phrase. And that got me thinking, combine that with a vowel reduction system, and that's almost like a stress-based vowel sandhi system: Every vowel in a phrase takes its reduced form except the last, which bears stress.

So I thought about applying that with this vowel reduction system I thought of a while ago: (S- strong/stressed, W - weak/unstressed/reduced)

Code: Select all

S  W

i  i
e  i
a  ə
o  u
u  u
ə  ə
aj e
aw o

So if you had a noun phrase made up of some words like /kaprau̯me do osimula/ it would look like: [kəpromidusimuˈlaː] (with likely contraction of the do and lengthening of the stressed vowel).

Now, that's where the sensible part of the idea ends.

I also had a very silly idea recently of making a languages whose lexicon consisted of just 80-120 CV monosyllables and nothing more. Not an oligosynthetic conlang - there would be homophones, and I would just force the language to deal with it. Oh, and no concatenating morphology. Maybe some genitive phrases like 'X of Y' to reduce ambiguity (proto-compounds, essentially, not yet lexicalized).

Then I thought, oh, what if I combined this with the vowel-reduction sandhi idea? Then I could have whole sentences consisting of nothing but stuff like /ki po ta vai̯ nu lo na nau̯ sa ne/ which gets broken up into maybe two syntactic phrases like: [kiˈpoː | təvenulunənoˈsaː | ˈneː ‖] (no idea what the actual syntactic structure would be there)

But then I remembered this would necessarily be very syntax-heavy, and that I hate syntax (well, kind of). So to at least take some of the burden off of the syntax, I would want some apophonic inflectional morphology. The vowels aren't a very good choice, since a bunch of the contrasts get erased in non-final position. So it must be the consonants. But that would decrease the number of distinct lexical items if I have to dedicate some consonant contrasts to apophonic inflection. So basically I can either abandon the strict CV template, or use a fairly large consonant inventory to ensure I have plenty of possible contrasts for both inflection and the lexicon. This would also detract from the goal of only having 80-120 possible syllables.

I don't really want to use tone, since the whole idea of the vowel sandhi thing was to do a tone thing without tone, but maybe perhaps a prominent lexical category, like person/number on verbs, could bear a floating tone that migrates to the stressed final syllable of the verb phrase. Or is distributed across the verb phrase. If there is a verb phrase? Not all languages have verb phrases. So I don't know. I also thought about maybe introducing phonemic gemination, but I almost feel that would ruin the rapid-fire mumbly rhythm of the language (as I imagine it).

Any other ideas for this crazy scheme are welcome.

[Frislander]

/p t t͡s k/
/s h/
/m n ŋ/
/w j/

/i iː o oː/
/eː a aː/

Word-internal consonant clusters are restricted to /s h/ + /p t k/, while word-final consonants are restricted to /k ŋ w/, where /w/ only appears as part of the diphthongs /iw aw/.

[Frislander]

/p t k/
/h/
/m n ŋ/
/w/

/i u/
/e o/
/a/
/iu ui eo oe/
/ai au/

Syllable structure is (C)V(K), where onsetless syllables are only found word-initially and K is restricted to word-final /k ŋ/.

[CarsonDaConlanger]

/m n/
/p t d ts tɬ k ʔ/
/ɓ ɗ k'/
/s ɬ x/
/l j w/

/i u/
/e o/
/æ ɑ/

+vowel length

CV(V/R)
R=l,n,m

How realistic do you think this is?

[sangi39]

/m n/
/p t d ts tɬ k ʔ/
/ɓ ɗ k'/
/s ɬ x/
/l j w/

/i u/
/e o/
/æ ɑ/

+vowel length

CV(V/R)
R=l,n,m

How realistic do you think this is?

I really like the /ɓ ɗ kʼ/ which made me think of Proto-Mayan to begin with (which has /ɓ tʼ kʼ/ and so on, and the singular voiced /d/ reminds me of Finnish.

The only thing that's giving me some pause for thought is having /ts tɬ/, /s ɬ/, ejectives, but no /tsʼ tɬʼ/. I can't think of a language off the top of my head that has that sort of gap (where a language lacks /tsʼ tɬʼ/ it tends to lack ejectives entirely, IIRC). I'm sure someone can point out a counterexample, but from what I can tell, affricates tend to pattern as plosives, so they tend to make the same distinctions, e.g. if plosives can be voiced, affricates also tend to have voiceless vs. voiced pairs, if there are ejectives, affricates also tend to have ejective counterparts, and so on.

[cedh]

I agree with sangi39 that the lack of /tsʼ tɬʼ/ is fairly unusual in such an inventory, but given your relatively simple syllable structure, this lack is easy to explain with diachronics: Suppose an earlier stage of the language had no glottalized consonants, with only /p t k ʔ/ as voiceless stops, but allowing them in coda position. Then the following sound changes occurred (among others, presumably):

{ps ts ks} {pɬ tɬ kɬ} > ts tɬ
p t k > ʔ / _C, _#
ʔ > zero / _CC (this eliminates any ʔts ʔtɬ clusters; alternatively the outcome could be a preceding long vowel)
ʔp ʔt ʔk > pʼ tʼ kʼ
Vʔ > Vː / except _V
pʼ tʼ > ɓ ɗ

and voilà, you get basically the described phonological distribution.

[Frislander]

I think the only place I've seen a language with both ejectives and affricates but no ejective affricates is actually Na‘vi, a conlang. If anything the tendency is the other way round - languages with both ejectives and affricates tend to be more likely to have only ejective affricates than non-ejective ones (see the Salishan family, which universally has /t͡ɬ’/ but almost as universally lacks its non-ejective counterpart), as well as many others such as Sandawe in Africa.

[CarsonDaConlanger]

I've removed /ts/ entirely, and I've got diachronics where older /tɬʼ kxʼ/ > /tʼ kʼ/ > /kʼ kʼ/

[Porphyrogenitos]

/m n ɲ ŋ/
/p t c k ʔ/

/a i u/

Syllable structure is (C)V(C). Nasals assimilate in place to a following consonant, non-nasals assimilate totally to a following consonant. Stops are voiced after nasals, and non-glottal stops are lenited and possibly voiced between stops, maybe something like [ɸ θ ç x] or [β̞ ð̞ j ɰ]. Word-final stops are likely unreleased.

Practically speaking, this means the following segments are contrastive intervocalically:
[m n ɲ ŋ]
[mb nd ɲɟ ŋg]
[pp tt cc kk]
[ɸ θ ç x]

Gemination may also be possibly phonemic, or conditioned at a fairly deep level, due to stress-related phenomena.

There is a process of rightward frontness harmony: After a syllable with /i/, all instances of /u/ become /i/. (The reverse is not the case - in this way, /i/ is "dominant" and /u/ is "recessive".)

There may be remnants of a historical voiced/lenis series - I'm thinking that perhaps a series /b d (ɟ) g/ went to zero postvocalically and merged with the voiceless series word-initially and in geminates, with the distinction only resurfacing in patterns of sandhi or morphological gradation.

The palatal series may be secondary in origin, reflected in synchronic patterns of palatalization. /ŋ/ may also be only recently phonemic from /ng/ clusters; /ʔ/ may also be recently phonemic from lexicalized sandhi effects in compounds or morphology.

[Zekoslav]

It's not quite a phonology idea, but I've come up with the idea of a descendant of Koine Greek (spoken somewhere in the Middle East) where φ, θ and χ didn't become fricatives.

Instead, there would be a chain shift where π, τ, κ > /b/, /d/, /g/ and φ, θ, χ > /p/, /t/, /k/. A little bit later, β and γ would devoice word-initially becoming [f] and [x] while remaining [v] and [ɣ] word-internally. This is simply a continuation of tendencies already present in Koine Greek (apparently the voiced stops became fricatives earlier than the voiceless aspirated ones, and there was positional voicing of voiceless stops) and would result in some serious obstructions to mutual intelligibility (the mismatch between consonants would be like that between Ancient Greek and Ancient Macedonian), especially if there was some serious vowel reduction as well!

[Creyeditor]

I really like that idea. Did you already think of other changes? Maybe one could do something nice with the vowels (keeping half of them from becoming /i/)?

[Zekoslav]

I really like that idea. Did you already think of other changes? Maybe one could do something nice with the vowels (keeping half of them from becoming /i/)?

During the relevant period, only ει and ηι had merged with ι, so you can definitely avoid rampant iotacism. It's likely too late to preserve vowel length, though: the language had stress accent and incipient vowel reduction. With the initial inventory of /i y e ε a o u/ that could lead to some interesting developments (maybe even some sort of consonant mutation, given that final /n/ would likely be lost as in IRL Greek).

The only particular change I've come up with is related to the change /u/ > /v/ when in diphthongs. In IRL Greek it eventually devoiced to /f/ in front of voiceless consonants. In this language it would be the opposite (maybe even the trigger for the general voicing of voiceless stops), so that e.g. αὐτός would become /avˈdos/ (and, after vowel reduction, who knows what...)

[Frislander]

/p t kʷ q (qʷ)/
/p’ t’ t͡ɬ’ t͡ʃ’ kʷ’ ʔ (ʔʷ)/
/s ɬ ʃ xʷ ʜ (ʜʷ) h/
/m n j w/

/i ə~u/
/ɛ a/

The /ə~u/ contrast on the surface is restricted to appearing before uvulars a /q ʔ ʜ/, or alternatively you could interpret this as a labialisation contrast which is otherwise not seen on these consonants, as evidenced by the presence of /u/ before /kʷ kʷ’ xʷ w/ in the same manner. Similarly there is a collapse of the /i ə/ and /ɛ a/ contrasts before palatals, with the front vowels only being found in those contexts.

Syllable structure is CV(C)(C), where any single consonant can appear in the coda, while clusters are restricted to /s ɬ ʃ xʷ ʜ (ʜʷ) m n/ plus a stop, and /p t ʃ kʷ q (qʷ)/ + /s ɬ ʃ xʷ ʜ (ʜʷ)/. These are the only clusters permitted - the morphology is structured to prevent other clusters from arising.

[LinguoFranco]

/m n ɲ/
/p t c k ʔ/
/mb ⁿd ŋɟ ŋg/
/s z f h/
/l ʎ r/
/j w/
/t͡ɕ d͡ʑ/

/i iː u uː/
/ɛ eː ɔ oː/
/a aː/

/mb ⁿd ŋɟ ŋg/ are supposed to be prenasalized consononats, but the IPA Keyboard only has /ⁿ/. I don't have much for phonotactics yet, although it is looking somewhat similar to my main project, with length distinction on vowels, and will likely be mora-timed. However, I think this language will have palatalization, so /ⁿdʲ/ would become /ⁿd͡ʑ/, for example.