Re: Kijeb sandhi

Replying to:
  Roger Mills <rfmilly@...>,
  Roger Mills <rfmilly@...>,
  Alex Fink <a4pq1injbok_0@...>


Roger Mills <rfmilly@...> wrote:

> BP Jonsson wrote:
> > I have entered some preliminary stuff for a new and
> > expanded section on Kijeb sandhi to
> > <http://wiki.frath.net/Talk:Kijeb>, > I'd like your
> > input on how realistic (phonetically plausible) the
> > sandhi changes look. Any suggestions that may make them
> > more realistic or interesting are welcome!
> Not dissimilar from some of the changes I needed in Gwr,
> for which I also had to make a chart (if only to keep
> things straight in my head...)-- the difference is that I
> used C-clusters to derive new single C.
> Everything looks quite realistic, and/or explicable. I see
> that initial *y, uniquely, metathesizes with C2; a more
> general application might also metathesize *w.
But that is what happens! Consider what happens with the
coronals and velars:

     wt > tw wd > dw wn > nw wk > kw wg > gw wñ > ñw

The combinations _wr, wy_ remain, but when labials are
involved a rule which deletes _w_ after labials comes
into play.

> Could things like ty, dy, sy, ky, gy etc. ult. > unit
> afficates?
Yes they do:

     ty > ts; > ts > s
     dy > dz; > dz > z sy > hy > h ky > cC > ts\ gy > J\j >
          dz\ ( > z\ > j )

There is also s > h / V_, so the new s only partly merges
with the old. Some dialects also have d > D > z / V_, so /z/
which didn't exist in Kijeb ([z] being only an allophone of
/s/ before /b d g/) becomes rather prevalent.


> In Gwr, clusters of pr, br > t, d-- not sure how realistic
> that is, but it's a sound change claimed by Benedict in
> his Austro-Thai work.
Well pr, br > retroflex t`, d` in most of Tibetan.

> Similarly, mr,nr,Nr develop an excrescent stop (mbr, ndr,
> Ngr), then the nasal is lost and we get br > d, dr >
> [d`l`] and [dZ].
I've got stop excresscence in these in Sohlob too, but no
nasal loss.

> Have you tried writing up the rules for your changes? That
> can sometimes be illuminating, in that you may see other
> possible generalizations.
I started writing the rules, then I realized I needed a
table for a birds-eye view. I actually saw the
generalizations better that way, like how regressive voicing
assimilation and t, d, n > r / C_ interact.

Roger Mills <rfmilly@...> wrote:


> BP Jonsson wrote:
> > BTW I'm at a loss how to express the various levels of
> > development in plain text. On the wiki I use italics for
> > Kijeb and bold face for the daughter languages. Using
> > *word* for the daughter languages and _word_ for Kijeb
> > kind of feels wrong, since it is Kijeb which is
> > intrafictionally largely reconstructed, since its
> > writing system is defective, and now Pre-Kijeb enters
> > the picture! Maybe I should change the notation to:
> >
> > :                 plaintext wiki
> > :
> > :  daughters     _word_ bold
> > :   Kijeb   *word italics
> > :   attested K.   _WORD_    SMALL-CAPS
> > :   reconstr. K.    **word *italics
> > :   Pre-Kijeb     **word** **italics
>
> Not sure I understand the exact sequencing,
It's like this, although a wave model might illustrate the
relations between the nodes lower than Kijeb better:

:               *Pre-Kijeb
:                   |
:                 Kijeb
:                   |
:      +------------+
:      |            |
:      |            +----------+---------+
:      |            |          |         +----------+
:      |            |      Classical     |          |
:      |            |       Sohlob       |          |
:    Linjeb      Kidilib               Heleb     Southern
:                                                dialect

Of these only Pre-Kijeb is entirely unattested. Kijeb is
attested in a defective script (a syllabary not
distinguishing between voiced and voiceless stops, nor
between stops and fricatives other than <S>) and with a
limited vocabulary. All the others are amply attested,
although Heleb and Southern are not normally written, using
Classical as their Ausbausprache. Classical itself is
nobodys native language, but an archaizing and somewhat
artificial literaty and ceremonial language, among other
things containing partially or wrongly adapted loanwords
from Kidilib and Heleb which don't follow the sound laws.

Extrafictionally this is because Classical retains some
words that would be derived from forms that aren't licit in
the current version of Kijeb, like the word _Kijeb_ itself:
if a Pre-Kijeb **kikipa had existed it would under the
present rules dissimilate to Kijeb *tikipa or *kitipa, which
would give _*tigeb_ or _*kideb_ in Classical. However *ti
and *ki merge as initial /ts\/ and medial /dz\/ in Kidilib,
so attested _Kijeb_ is a mixture of unattested 'pure'
Classical and Kidilib forms -- and so is _Kidilib_ of
course, which is _Tidilib_ /ts\idz\ilib/ in real Kidilib!
(Hint: _Kijeb_ means 'ancient language', _Heleb_ means 'city
language' and _Kidilib/Tidilib_ means 'language of the
ancient city'!)

Probably only Kidilib is actually descended from attested
Kijeb, the others being descended from closely related
dialects. Heleb and Southern have in common some vowel
developments not shared by Classical and Kidilib, but then
Southern has consonantal developments all its own. Linjeb is
sui generis in both consonantism and vocalism, and most
conservative in not partaking in consonant lenition or vowel
harmony, but it shares some of the ways it treats
palatalized consonants with Kidilib (in general they
palatalize more strongly than the others).

An important isogloss is the treatment of nasal+stop
clusters *mp, *mb, *nt, *nd, *ñk, *ñg:

*     Linjeb: unchanged
*     Kidilib: _hm, mb, hn, nd, hng, ng_
*     Classical: _mb, mb, nd, nd, ng, ng_
*     Heleb & South: _mb, m, nd, n, ngg, ng_

The rationalization behind the Kidilib reflexes is that the
voiceless stop clusters went through a development like:
[nt_h] > nth > nh > [n_0] > hn. I'm not sure it is
realistic... The Classical output may be an artifact of an
archaic spelling which didn't distinguish voicing in stops
(normal native Sohlob spelling still doesn't, since voiced
and voiceless stops contrast only word initially, and
context normally disambiguates). Southern later goes through
some nasal harmony business, but the developments presuppose
Pre-Southern to have been like Heleb. (And yes, _Linjeb_
means 'western language', but I haven't decided the word for
'south' yet -- don't like the old one!)

>  but how about using labels:
>  * K word (attested Kijeb, if that level exist)
>  * PK proto-Kijeb
>  * PK *word
>  * Pre-K or maybe UrK for pre-Kijeb
>  * Pre/UrK *word
>  * A,B,C for various daughters,
>  * A word, B word etc.
Yes, I am using labels more or less like that ATM, but I
feel the need for some shorthand for Pre-Kijeb, and **word à
la the Nostraticists suggests itself.

FWIW _Kijeb_ is short for _Kij Sohlob_ 'Ancient Sohlob', so
it would be Proto-Sohlob rather than Proto-Kijeb. However
Pre-Kijeb is as far back as I go ATM -- I just got there in
fact, and I think the Sohloçan themselves went even that far
only exceptionally.

> (if you need reconstructed stages for A,B,C, then you
> could have "PA *word > A word", maybe even "PAB *word > A
> word1, B word2" etc. etc. Any intermediate hypothetical
> stages, use **, perhaps "pre-A **word > A word"
>  That IMO would eliminate some of the typographical
>  fussiness, and make the various levels clearer. Then use
>  italics just for glosses.
Yes, I think I'll use a combination of labels and asterisks
as you suggest, but keep the convention of writing the
daughter languages bold and older stages italic, together
with the labels.

Alex Fink <a4pq1injbok_0@...> wrote:


> On Thu, 29 Mar 2007 10:45:32 +0200, Benct Philip Jonsson
> <conlang@...> wrote:
> I have entered some preliminary stuff for a new and
> expanded section on Kijeb sandhi to
> <http://wiki.frath.net/Talk:Kijeb>, since I in the course
> of revising Kijeb verbal morphology I've found that I need
> [...] I'd like your input on how realistic (phonetically
> plausible) the sandhi changes look. Any suggestions that
> may make them more realistic or interesting are welcome!
> I like it! Looks realistic overall, but there are a few
> of what appear to be niggly exceptions; and that's
> perhaps one disadvantage of giving only a table, that
> isolated exceptions don't necessarily stand out. But the
> table is certainly a good idea for getting an overview of
> the whole system of sandhi. And you've named all the
> table cells, nice!
Yes, the names are helpful when hand-editing the table. FYI
I used a Perl script to generate the table before filling it
in, using an array containing the input phonemes and going
through it with two nested for loops.

> I'd expect pN bN > Nwgw. There's no reason for the
> sequence Nwgw to be excluded, given that labiovelars are
> unitary -- or aren't they?
You are absolutely right, except that the whole thing can be
written _ñgw_ as velar+labiovelar isn't possible anyway.
I've changed that. Thanks!

> The whole velar + labial corner is a bit anomalous as
> well, but I think it's what one would expect if velar +
> labial clusters collapsed to a labiovelar with manner and
> voicing taken from the velar before most of these changes
> happened (but after nasal assimilation. And I'm assuming
> here an xw > f merger, which seems more plausible than x
> being redundantly rounded.)
You are absolutely right on all accounts.

> I'd also expect fm > m, since fric+nasal > nasal+stop
> didn't happen elsewhere.
> w seems to be metathesizing before non-approximants [w r
> j], except that for whatever reason in [wn ws] it doesn't
> make it all the way to [nw sw] but collapses into a
> single labial, which is odd given that [nw sw] are
> licit. What's your reasoning there?
A brainfart due to the fact that [nw sw] arent licit word
initially, but only word medially. I've corrected it.

> I can't see through your use of ny for the outcome of NN,
> but suspect there's probably a reason for it.
I think you mean Ny > ny. I assume that underlying /Nj/ and
/nj/ merge as phonetic [J], since I actually am unable to
distinguish [J] and [n;] by ear.

> Merging cells md and mn would be more parallel to the
> other nasals.
Done.

> Given that [z] is an allophone of /s/, is there really
> a sandhi change in [sb sd sg], i.e. should the cells
> be bolded?
True, but since /z/ was phonemic in the daughter languages
Sohloçan philologists transcribed Kijeb /s/ as _z_ where it
corresponded to /z/ in the languages they spoke. They had no
concept of phonemes and allophones, you see.

> Alex
Thanks for your keen observations. I added those and some
other corrections, and also a column for what happened to
word-final consonants.

/BP 8^)>
--
Benct Philip Jonsson -- melroch at melroch dot se

    a shprakh iz a dialekt mit an armey un flot

                                 (Max Weinreich)