Hallo!

On Sun, 28 Feb 2010 22:07:56 +0100, Benct Philip Jonsson wrote:

> On 2010-02-28 Alex Fink wrote:
>  > Interesting.  I suppose historically you imagine an /i u a/
>  > system, followed by lots of colouring (not that I see how that
>  > fits in with PIE).  How does the colouring become harmonic,
>  > then -- how does it come to pass that the same colourant vowel
>  > affects all the others, and then (presumably) disappears?  Or
>  > have I guessed wrong, and there's a more complex story?
>  >
> 
> That's exactly the situation I have in Sohlob.  Pre-Sohlob,
> and the oldest attested language Kijeb, had /a i u/ with
> subphonemic 'coloring' by following vowels as follows:
> 
>          /a/     /i/     /u/     <-- Vowel of following syllable
> /a/     [a]     [&]     [Q]
> 
> /i/     [e]     [i]     [i\]
> 
> /u/     [o]     [u\]    [u]

This is almost exactly what happened in early Old Albic!  Except that
/i/.../u/ probably gave [y] rather than [i\].

The further developments depend on dialect.  In the southern dialects
on which Classical Old Albic is based, [Q] merged with [o], [&] merged
with [e], and [u\] merged with [y].  Another source of mid vowels were
/aja/ > /e:/ and /awa/ > /o:/, with subsequent shortening of the
resulting long vowels in closed syllables.  The groups /awe/ and /ajo/ 
(resultung from i-umlaut of /awa/ and u-umlaut of /aja/, respectively)
gave /2(:)/.  Further long vowels resulted from the loss of certain 
consonants, most notably /3/ (which probably was a pharyngeal
approximant [?\].  The result is the classical Old Albic vowel system

/i(:)/ /y(:)/        /u(:)/
/e(:)/ /2(:)/        /o(:)/
              /a(:)/

The developments in other dialects are mostly not yet explored.

> Next vowels in unstressed syllables become [@], some of
> them eventually disappearing entirely, making the
> various vowel allophones in the preceding stressed
> syllable phonemic.  Basically the first syllable of
> every word had the main stress and very second syllable
> thereafter had secondary stress with the exceptions
> that
> 
> 1) the first syllable of each element of transparent
>     compounds received main stress, starting a new
>     'count' of alternating syllables and relegating the
>     preceding syllable to unstressed, unless it was
>     word-initial.
> 2) the last syllable of polysyllables was always
>     unstressed unless it constituted a monosyllabic
>     compound element (rare).
> 3) The penultimate syllable of polysyllables was always
>     at least secondary stressed.

Interesting stress rules; I like them.

> Later /o/ merged with /Q/ or /u/, /e/ merged with
> /&/ or /i\/ depending on dialect and /u\/
> merged with  /i\/, giving the following system:
> 
>      /i/     /i\/    /u/
> 
>      /&/     /a/     /Q/
> 
> Lastly the vowels of each word became all high or all
> low depending on whether the stressed initial syllable
> being high or low.  Thus vowel harmony.
> 
> This scentario is some ten years old.  I'm not sure anymore
> that it's naturalistic.

It is naturalistic, I think.

>       I'm even less sure about the
> scenario by which some dialects are supposed to later have
> developed front-back harmony overlapping with the high-low
> harmony, namely that
> 
> a) VCV sequences where the C was one of /j w G h/ became
>      one long vowel.

Similar to what happened in Old Albic (see above): VCV with
C being one of /j w 3/ become V: with /j/ contributing the
feature [+front] and /w/ contributing the feature [+round];
in closed syllables, the vowels was later shortened.

> b) The front/back-rounded/unrounded value of these long
>      vowels was a combination of the values of the members
>      of the original VCV sequence as follows:
> 
>      i) If the original combination contained one of
>          /i & j/ the resulting vowel became front, else
>          it became back.
>      ii) If the original combination contained one of
>          /u Q w/ the resulting vowel became rounded, else
>          it became unrounded.
> c) Long vowels, or the first long vowel of a word if there
>      were more than one, attracted the stress.
> d) All vowels of a word got the same frontness
>      value as the stressed vowel. Thus the only still
>      distinctive feature of unstressed vowels was roundness.

This all makes perfect sense to me.

> Thus there were in stressed syllables eight distinct
> vowel qualities plus a length contrast
> 
> /i(:)/  /y(:)/      /M(:)/  /u(:)/
> 
> /&(:)/  /2(:)/      /A(:)/  /Q(:)/

The Turkish vowel cube, essentially!  I always liked that system.

> There are the same eight qualities occurring
> in unstressed syllables, but all features except
> roundness -- i.e. height and frontness -- are
> determined by harmony with the stressed vowel!
> 
> (One branch of the family never developed VH, rather it
> differed from the others so that instead of the system
> of nine vowel allophones it had eight allophones, so
> that both /i/:/u/ and /u/:/i/ were [y]. After
> phonemization of the allophones by reduction and loss
> of unstressed vowels as in the other dialects it went
> through a push chain /Q/ > /o/ > /u/ > /u\/, with later
> /u\/ > /8/ > /2/ and additionally /&/ > /e/, resulting
> in the non-harmonic, assymetric vowel system
> 
>      /i/     /y/                     /u/
>          /e/     /2/             /o/
>                          /a/
> 
> This is a revision from an earlier scenario where the push
> chain was /Q/ > /o/ > /u/ > /y/ > /2/, which probably is
> unnaturalistic.

The u-turn it makes looks indeed a bit weird.

>       I'm now suspecting that /u\/ > /8/ > /2/
> is unnaturalistic as well, or at least uncomfortably close
> to what actually happened in my L1! :-)

If it happened in your L1, it isn't all that unnaturalistic, but if
you don't like it, change it.

... brought to you by the Weeping Elf
http://www.joerg-rhiemeier.de/Conlang/index.html