Decircumflexion, N-raising, H-raising: Slavic soundrules on-line

From: Miguel Carrasquer
Message: 32221
Date: 2004-04-24

On Wed, 21 Apr 2004 21:32:33 +0200, Miguel Carrasquer
<mcv@...> wrote:

>On Wed, 21 Apr 2004 19:08:42 +0000, Sergejus Tarasovas
><S.Tarasovas@...> wrote:
>
>> *-u:m > *-um looks like a regular Slavic development
>>(do you have any objections?)
>
>I couldn't believe it at first. Looks like a stroke of
>genius. I'm not yet 100% sure it doesn't mess up anything
>somewhere else (acc.sg. *-eh2m > -á:m > -ó:m > -om > -oN [or
>without nasal raising: -á:m > -am > -om > -oN] is safe). I
>should do a complete check with the Sound Change applier.
>Later.

Well, after struggling the whole day to get it working, I've
finally succeeded.

The idea is simple. There are three vowel raising factors:
(1) circumflex raising [C]
(2) *-n or *-m (> *-N) raising [N]
(3) *-s (> *-h) raising [H]

There are three factors affecting the length:
(1) shortening by decircumflexion [DC]
(2) shortening by elimination of long diphthongs (V:i, V:u,
V:N, V:l, V:r) [LDE]
(3) lengthening by *-Rs [NH]

Frontness is affected by one factor:
(1) j- [J]

For the forms without j-, the picture is clear, e.g. for a
few sample endings:

C N H / DC LDE NH result
*-a::m 1 1 / 1 1 -u(N) f.gen.pl. -U
*-o::m 1 1 / 1 1 -u(N) m.gen.pl. -U
*-a::s 1 1 / 1 -u:(h) f.gen.sg. -y
*-o::is 1 1 / 1 1 -1 -u:(ih) m.ins.pl. -y
*-o::i 1 / 1 1 -ui m.dat.sg. -ô > -u
*-a:m 1 / 1 -oN f.acc.sg. -oN
*-om 1 / -u(N) m.acc.sg. -U
*-os 1 / -u(h) m.nom.sg. -U
*-a:ms 1 1 / 1 -1 -u:(Nh) f.acc.pl. -y
*-o:ms 1 1 / 1 -1 -u:(Nh) m.acc.pl. -y

The problems start with the "soft" endings, which have f.
gen.sg. (and f.nom.pl.), m/f/ acc.pl. -jeN where the "hard"
endings have -y. After having hacked through most
permutations in the ordering of C-raising, N-raising,
H-raising, shortening, lengthening, and J-umlaut, and never
getting it right, I accidentally (well, by quitting hacking
the code and thinking about the matter) stumbled upon the
corect answer (at least, an answer that gets all the results
right in the Sound Changer).

Surprisingly, this answer does not involve the Umlaut of
-jo- > -je-, as I had incorrectly assumed all day long. The
answer involves raising *-o:ms/*-a:ms all the way to *-u:Nh,
and then, to quote the Sound Changer:

[input: g^hem-j-eh2-ms]
...
a:N->o:N /_(h)# applies to 3emja:Nh at 8
o:->u: /_(N)h# applies to 3emjo:Nh at 8
u->i /j_ applies to 3emju:Nh at 5
i:->X /_ applies to 3emji:Nh at 6
X->i /_ applies to 3emjXNh at 5
3->z /_ applies to 3emjiNh at 1
iN->e~ /_ applies to zemjiNh at 6
h-> /_# applies to zemje~h at 7
[output: zemje~]

The only trick is the development -iN > -e~, besides -eN and
-IN > -e~ (this part *was* accidental, I meant -IN > e~, but
the code said -iN).


The full set of input forms and rules, now covering most of
PIE > Slavic (doubtlessly with a large number of bugs) can
be seen/played with at
<http://home.planet.nl/~mcv/PIE_Slavic.html>. I should've
checked the etymologies of krajI and lêto before putting it
online, but never mind...


=======================
Miguel Carrasquer Vidal
mcv@...