] On 16 April 2017 at 16:02, Melroch <[log in to unmask]> wrote:
>> I'm tinkering with Celimine again, and thought it might be nice to get
>> some feedback on the broad outlines of its diachronic phonology.
>> The consonants in Proto-C have a palatalized / velarized distinction,
>> similar to Irish or Russian (although not every consonants in the
>> "palatalized" is actually phonetically palatalized, and not every
>> consonant in the velarized set is actually phonetically velarized).
>> The inventory as currently envisioned is as follows:
>> Palatal  ~  Velar
>> Plosives:
>> pʲ  bʲ ~ p  b
>> tʲ  dʲ ~ t  d
>> c ɟ ~ k  g
>> Nasals:
>> mʲ ~ m
>> ɲ ~ n
>> ŋʲ ~ ŋ
> Thats a pretty tenuous distinction ɳ vs ŋʲ you have there.

? I don't have a retroflex nasal ɳ there...

> I'd expect nʲ
> vs. ɲ because ŋʲ pretty much is another, usually phonologically motivated,
> way of writing ɲ.

Well, yeah. Doesn't make a difference to the actual content, but I
suppose notating the palatalized version of /n/ as nʲ does make sense
given the regularity of the pairing throughout the consonant

> It would be a nice touch if there was no coronal/dorsal
> distinction in nasals on the palatal side though. This phonology cries out
> for an assymetry. :-)

Why do you say that? Just because it is exceptionally symmetrical
right now? Incidentally, /nʲ/ and /ŋʲ/ are neutralized under lenition
(both become just /j/), so just eliminating the distinction entirely
would kinda fit.

>> Fricatives:
>> fʲ  vʲ ~ f  v
>> z ~ ʂ
>> Approximants:
>> ɹʲ ~ ɹ
> ɹ ~ r/ɾ would be kind of realistic here.

That seems reasonable.

>> l ~ ɫ
>> I figure at an even earlier stage, /ʂ/ developed from an unvoiced
>> version of /ɹ/, and then got re-analyzed to pair up with its fellow
>> sibilant /z/ instead. Or perhaps it came about as a merger of /ɹ/ and
>> /z/, or something like that.
> Any particular reason it's not just s?
> In fact sʲ becoming ʂ and s becoming
> z seems more likely.

Reverse-engineering. sʲ -> ʂ is indeed more likely, but in order to
get the right distributions of [s] vs.[ʂ] allophones in the modern
language, I need an excuse for ʂ to pattern with the velar consonants
and trigger vowel backing. By itself, that's not implausible,as such
patterning is attested in natlangs (Russian treats ʂ as a "hard",
velarized consonant), but deriving it from palatalized [sʲ] would
require it to group the other way!

>> So, how does that look as far as plausibility for a natural phonemic
> OK, except for the unlikely z ~ ʂ pairing.

Any better if I have separate zʲ ~ z & s ~ ʂ pairs? Or just replace it
with  s ~ ʂ?

>> The sound changes that produce the modern Celimine phonemic inventory
>> involve a process of displaced contrast that collapses the
>> palatal/velar distinction in consonants and splits the vowels into an
>> 8-vowel system with front/back harmony, and a process of intervocalic
>> lenition that introduces a bunch of new fricatives and approximants,
> Shame on you for rushing over the most interesting part! ;-)

Well, the main thing is just to see how the proto-phonology works on
its own. What it turns into *later* is just background....

I'm still not quite clear on the exact order in which all of the sound
changes need to occur (I wanted to get feedback on the proto-inventory
before spending more effort on that), but if you are interested, the
general trends are as follows:

Vowel breaking occurs as you would expect.

pʲ/bʲ -> p/b
fʲ/vʲ -> f/v
tʲ/dʲ collapse with t/d phonemically, but maintain a phonetic
distinction that ends up getting generalized so that contact with
different classes of vowels due to morphological processes, etc.,
trigger consonantal allophony independent of the historical palatal
vs. non-palatal status of a synchronic /t/ or /d/.
c/ɟ do pretty much the same thing with k/g

mʲ ~ m
ɲ ~ n
ŋʲ ~ ŋ

z ~ ʂ

ɹʲ ~ ɹ
l ~ ɫ

Intervocalic lenition (for which purposes r and l also count as vowels) turns:
tʲ, dʲ, z, & ʂ into
t/d into θ/ð (so there is a modern alternation between θ/ð and h)
p/b into f/v (both palatal and non-palatal varieties)
k/g into x/ɣ
m into w (both palatal and non-palatal varieties)
nʲ/ŋʲ int j

Simplification of geminates then restores single intervocalic stops to
the inventory (either the modern language still has some geminates, or
harmony-resolution rules need to come into play, depending on whether
this occurs before or after vowel breaking).

>> resulting in these modern inventories:
>> Vowels:
>> /i/ /u/
>> /ɪ/ /ʊ/
>> /e/ /o/
>> /ɛ/ /ɑ/
> I'd expect [æ] [ɑ]. Pretty crowded in front otherwise.

True. I'll have think about that. [æ] is such an Englishy vowel, though....

>> Stops:
>> p b
>> t d
>> k g
> I'd expect at lest some coronal affricates given all those ancestral
> palatalized stops. There is French -- and Swedish and Norwegian! -- for
> ANADEW though.

Oh, don't worry- as presaged above, you get those as allophones! /t/
and /d/ realize as /t͡ʃ/ and /d͡ʒ/ next to front vowels, precisely
because those front vowels were created by the presence of ancestral
palatal stops. Most of the phonetic palatal/velar consonant pairs
collapse during the vowel-splitting process, but not all of them.

>> Nasals:
>> m
>> n
>> ŋ
>> Fricatives:
>> f v
>> θ  ð
>> s z
>> x  ɣ
>> h
>> Approximants:
>> w
>> j
>> r
>> l
> What happened to ɲ? Merged with j?

Merged with /n/, except where it turned into /j/.