Re: Cwendaso: when two diphthongs collide

> > The most notable thing is that two like diphtongs coalesce into one instead
> > of the high vowel in the first vowel becoming a glide.
>
>You could explan this as haplology: the reduction of adjacent identical
>parts of a sentence. Otherwise, this looks good.
Thanks for the explanation and terminology.  I suspect that some other
dialects of the language may not reduce them to a single diphthong, but
I'll be dealing mainly with the dialect spoken in one particular village,
and that dialect does have pretty uniform haplology in V+V sequences.

>Except, that don't you have <ai-eu> for example, behaving differently from
><ai-e>?
Can't remember :-)  I'd really have to go back and look; I put down such a
mass of information that I can't remember it all.  Here, let me take a look
at what I put down in my notebook, since I'll never find what I want in my
mailbox quickly.  Here it is:

<ai-eu> --> <ayeu> and <ai-e> --> <ai>

>  That doesn't really seem right to me.
You're right; it seems odd.  I've been thinking of things like <ai-e> -->
<ai> in terms of the <e> being raised to assimilate with the preceeding <i>
and then being absorbed into it because the language doesn't have
geminates.  If a <Vi-e> sequence causes the <e> to rise, then we'd expect
<ai-eu> --> <ai-iu> --> <aiu> --> <ayu>  (which is also considerably
shorter and easier to pronounce than <ayeu>.)

>But maybe it can work. Here's a thought: perhaps syllable weight plays a
>part in vowel reduction?  Consider a hypothetical word like /kai-edu/,
>which IIRC reduces to [kai.du].
Correct, it does reduce to that.

>  Before reduction, we have /ka.ye.du/--so
>we can look at this as reducing the syllable /ye/ to /i/ by dropping the
>/e/.
Or by first raising the /e/ (assimilation) and then absorbing it (whatever
the proper tecnical term for that is -- deletion?) to prevent a geminate,
as I demonstrated in one of the above examples.  This raising followed by
deletion could account for the slightly abnormal behavior of V+ei
combinations, where that medial /e/ simply drops out and you get a V+i
diphthong.  But how would this effect the rest of the system?  Shouldn't
the other mid-vowel, /o/, then do likewise in the the /ou/ diphthongs?  But
this is exactly what is happening in <a-ou> --> <au>.  OTOH, I curently
have <e-ou> --> <you> from the /e/ rising to become a glide as it does
everywhere else.  I suppose this could simple be a matter or
rule-ordering.  But OTOH again, my notebook shows <pra-ois> -->
<prais>.  There seems to be something irregular about the behavior of /a/
combined with a diphthong.

And I just *now* realized that I completely forgot to run test cases of all
the vowels followed by /ai/!  Without using actual example words, here's
what I think they should do.

<i-ai> --> <yai>
<e-ai> --> <yai>
<a-ai> --> <ai>
<o-ai> --> <oi>
<u-ai> --> <wai>

The only doubtful member of this set is <o-ai> --> <oi>.  If /o/ rises to
[w], then it should be <o-ai> --> <wai>, but if the /a/ and /o/ phonemes
sort of "pull on each other", trying to assimilate to each other when they
end up next to each other (and another vowel is also involved), then that
could account for both <o-ai> --> <oi> and <pra-ois> --> <prais>.  (And if
this dialect, as I mention below, has something of an aversion to
overly-long strings of vowels and semi-vowels and attempts to reduce them,
it might choose to assimilate the two vowels to each other and then drop
the middle one.)

There's still the problem of the behavior of /eu/, which is a difficult
sound for me to produce.

<i-eu> --> <yeu>
or would the /e/ be likely to rise, giving us <yu>?
<e-eu> --> <eu> obviously
<a-eu> --> <ayu> if the /e/ rises.
If not, we could end up with the /e/ pulled downwards toward the /a/ and
get <au>, which seems less likely to me.  The /e/ seems to be inclined to rise.
<o-eu> --> <oyu>
<u-eu> --> <weu>

>  Now, in the word /kai-eudu/ we have /ka.yeu.du/ before reduction--but
>in this word reduction doesn't happen. That's because the middle syllable
>/yeu/ is heavy (has a coda), and so cannot be reduced.
What is the coda, the /u/?  (I guess it must be, since it has turned into a
glide, which is a type of consonant.  I had never been taught to think of
it in terms of such a syllable having a coda.  I was always told that a
syllable containing a diphthong was heavy, or "long" in the case of Latin,
and so was one with a vowel followed by two consonants.  If the second half
of the diphtong is a consont, then the syllable is closed, I guess.)

>  Perhaps this
>explains the different behavior of diphthongs?
Probably I'll go back over my data and reduce <ai-eu> to <ayu> and other
similar things instead.

>And you can still use haplology to account for the exceptions.
Which I probably will, at least in this dialect.  As I said in a previous
post, certain other dialects prefer to split final diphthongs into two
syllables, either that or pronounce a stressed diphthong extra long with a
falling intonation pattern on the final vowel that implies that it does not
really belong to the same syllable as the first portion of the
diphthong.  (Sorry, can't give you a sample of what I mean since I can't
transcribe the subtleties of intonation.)

This implies that something like /kai-ekka/ doesn't reduce to [kaik.ka],
>but remains [ka.jek.ka]. Essentially, it boils down to a prohibition of
>superheavy syllables, which contain both a diphthong and a coda consonant
I'll have to give careful consideration to a rule like this, in some form
or another, because I already sense that the language doesn't like the
codas to get too crowded (e.g. /rm/ is an illegal syllable coda, meaning
that /lm/ would be similarly illegal, and I have one example of /rd/ being
an illegal coda requiring the insertion of an epenthetic [a] to break up
the cluster.  By extension, this would make virtually any /r/+stop sequence
illegal in the coda and might make any liquid+stop sequence illegal in the
coda.)  But I have also sensed that the particular dialect that I am
working with is not overly fond of long sequences of vowels and semi-vowels
and will do what it can to condense them to a more reasonable
length.  (Other dialects don't have this hang-up.)

Something else that could be going on in this language instead of syllable
weight is stress.  Cwendaso has free stress, and which part (if any) of a
pair of combining diphthongs or vowel and diphthong receives stress could
possibly cause variant outcomes for certain of the combinations.  When I
was combining those vowels and diphthongs, I was working with actual
*words* and saying them aloud to myself repeatedly to try to hear and feel
where the combinations would naturally want to end up.  Those words did not
have identical stress patterns, and that may have affected how I felt that
the diphthongs should meld in certain situations.  (The orthography of the
language has the stressed syllable in each word maked, but I didn't mark it
in my examples because I hadn't found the way to do it yet from my
keyboard.  And I still can't do every word, because sometimes it is a
syllabic sonorant that receives the stress, and I am not yet fully Unicode
enabled -- and neither is everone on the list.)  So when I was prefixing
<apai> to <outhú>, it was under different stress conditions than in the
same set where I prefixed <koi> to <óufl>, and that same set contained a
third stress pattern because the prefix <nggéi-> unlike most prefixes has
it's own stress and therefore steals stress from the root.  So this could
have possibly caused some varied outcomes in that set.  I didn't think
about it until afterward.

>Do with this as you'd like.
I truly appreciate your responses to my post.  It's been great working
together with you to get these aspects of the phonology up and running
properly.  Now I need to go back over my data and your comments, think
everything through, then go through my notebook and correct the paradigms
where I have chosen to revise them.  I think that I do need to come up with
a cohesive set of phonological rules to account for all the variations so
that this dialect has good internal consistency.  (Though I may still allow
it to do a somewhat irregular thing here and there since natural languages
do that; I'm aiming at good internal consistency, not absolute, unbroken
regularity.)

Gotta run.  My daughter has to go to a skating lesson and a violin lesson
in very quick succession: the skating lesson runs from 4:30 to 5:30, and
violin is from 5:30 to 6:10!)

Isidora