Hello Piotr,
I'm quite sure that we have also ul~lu ur~ru as output of l. r. in
PAlb. so the derivation:
Alb *gjuhë 'tongue' < PIE *gl.sos (Sl.->PIE *galsos) is correct
For sure I will wait until you will have more time, however I will
post my feedback here because more clarifications are need to be
added to your previous observations related to my examnples.
I.
>> Marius wrote
> > The outputs for PIE *l. *r. are also ul~lu and ur~ru and not
ONLY li~il and ri~ir
> >
> > Examples:
> > 1. grurë 'wheat' < PAlb *gruna: < PIE g'r.:nom 'grain' ->so r.
> ru
> Piotr wrote :
> Wrong example. There was a laryngeal in *g^r.h2no-, and that
should have
> given Alb. <ar>, not <ri>, let alone <ru>. Before a syllabic
consonant
> *g^ would have developed into <d(h)>. Your etymology is evidently
faulty.
The situation is not so simple as you have presented above.
There are Albanian words containing sonorants in which palatal
velars k',g' passed to pure velars k,g at a very early stage. This
depalatization seems to be directly linked to the presence of a
sonorant in the root (doesn't matter if it is syllabic or not).
However the exact context of this depalatization is not enough
clarified.
Examples:
1. Alb. mjekër < PAlb *mekra (reshaped form) or *mekri < PIE
*smek^r. => the depalatization of *smek^ru affected the Lith too:
smãkras but not the Sanskrit : smasru
NOTE : this is the second example (after grurë where g^/r. > g/r.)
showing a depalatatization based on r. -> k^/r. > k/r
(see at Demiraj: Alb. mjekër < PIE *smek^r.)
(Even it wasn't r. but only r is not very important for this
discussion: we talk here only about 'a sonorant context')
2. Alb. gju 'knee' < PAlb *gluna < *g(a)nu-na < PIE *g^onu-
based on PIE *g^enu => even the intermediary phases are not clear
enough the root is obvius *g^enu and the depalatization is obvious
here too.
3. Alb. krye 'head' < PIE *kra:nja: < krasnja: < cognate of Grk.
kranion
(It's true that I strongly suspect for krye a linked with Rom.
creier Rom. variant crier 'brain' < Lat. cerebelum)
4. Alb. grurë 'wheat' < gruna: < *gr.:na: < PIE *g'r.:no-
so for grurë we can consider *gr.na: at a very early stage.
Note: the laringeal that you have talked about produced only the
lengthening of r. -> r.h2 > r.:
So seems that the rules are:
r.: > ru~ur
r. > ri~ir
l.: > lu~ul
l. > li~il
What examples with r. < ar do you have?
5. I think also that a good example of r. > ur is also
Alb. gurë 'stone' < PAlb gura: < PIE *gWr.-
zero-grade of PIE *gWer- 'mountain' :
see Dacian. Giri-dava Skt. giri 'mountain'.
Note-1: if you compare:
PIE *gWr.- > PAlb. *g-(w)-ur- > gurë with:
PIE *wl.k- > PAlb. *(w)-ul-k > ulk
PIE *gl.s- > EPAlb *g-ul-s (later *glus-) > gjuhë
'strange enough' the match of r. l. is perfect....so no need for
exceptions regarding ul.k
Note-2: Without to consider r. in this Albanian form, we will be
obliged to think to a delabialization of gW/+, kW/+ that normally
passed to c^,g^ in PAlb.
Note-3: In order to avoid the expected output g^,
Demiraj 'forgot' the e in its derivation proposing gwrH-i
(but 'simple' gWr will not give gur but only gr, (see kWr.mi >
krimb) for this reason he supposed an early gW > gw as 'workaround')
II.
> > 2. ul-zë 'elm' < PAlb *ulma < PIE l.mo- 'elm' ->so l.
> ul
>
> And what happened to the *m? And if its loss can be somehow
explained,
We have:
ms > s (see PAlb *memsa > Alb. mish )
ns > s (see peshë from Lat.pensum ;
also the inherited pesë 'five' < *pensë < *penc^e/penc^a)
nz > z (see Rom. mânz > PAlb mëz < LPAlb manza < EPAlb mandja
so I expect:
mz > z too.
> how will you rule out borrowing from Latin?
Because I don't know Albanian 'Latin derivations' in -zë < PAlb *-
dja
(I skip out pupãzã that is onomatopeic so could appear in any
language)
But of course I cannot rule out 100% this option.
III.
> > So gjuhë 'tongue' < PAlb *glusa(:) < *gulsa(:) < PIE *gl.sos is
> > regular.
>
> Tou haven't offered even one independent example of *l. > lu, so
what
> are you talking about?
I hope that the clarifications above show that the examples aboves
are valid ones: ul~lu, ur~ru is possible from l. and r.
Also the intervocalic s>h is hard to be rejected.
Best Regards,
Marius