Albanian (3)

3. The incredible adventures of *s

PIE *s disappeared from in many positions already in the earliest
history of Albanian. It was lost before nasals (*smek^ru- > mjekër) and
probably always before and after liquids (the proposed development of
initial *sr- > *str- > shtr- is based on doubtful evidence). Some of the
examples discussed recently on this list suggest the compensatory
lengthening of vowels before *-rs- and *-sr- (simplified to -r-), as in
*g^Hesr- > *3'e:r- > dorë 'hand' and *gWer(h3)snah2 > *3^e:rra > zorrë
'intestine'.

Contrary to what is sometimes claimed, I can't see any tangible evidence
for the phonologisation of the RUKI rule in Albanian.

Before Late Proto-Albanian (i.e. in pre-Roman times) *s underwent
voicing prevocalically in stressed syllables but remained voiceless
elsewhere (unless already lost). After the main wave of Balkan Latin
loans in Albanian but before the main wave of borrowings from Slavic the
fricatives *s and *z became retracted, yielding postalveolar *s^ and
*z^. While *s^ survives as Modern Albanian <sh>, *z^, which had
meanwhile undergone a merger with the reflex of Proto-Albanian *j,
developed further into palatal *G' and merged with *g' (from *gl or from
*g palatalised before front vowels). The outcome of the merger is Modern
Albanian /g'/ (spelt <gj>):

*serpm.no- > *ziärpan- > Geg gjarpën, Tosk gjarpër 'snake'
*supno- > *zum(n)- > gjumë 'sleep'
*solwo- 'healthy' > gjallë 'vigorous'
but *septm. (with final stress preserved) + *-t-a: > *s&táta > shtatë

Alb. <sh> normally corresponds to Lat. /s/ in old borrowings (e.g.
shigjetë 'arrow' <-- sagitta), but not to Slavic /s/, except in several
words that belong to the oldest layer of Slavic loans (e.g. shullë
'maidservant' <-- *sUlU 'messenger').

Combinations of *s with other phonemes have produced some really bizarre
results. Initial *sw- gives Mod.Alb. d-, as in *sworgH-eje- > dergjem
'lie ill' (middle voice) and *s(h2)wel-jo- (or the like) > diell 'sun'.
The development is puzzling and the intermediate stages are difficult to
reconstruct. The syllables in which it took place were stressed in
Proto-Albanian (though not necessarily in PIE), so we may conjecture
that the first step was *sw- > *zw-. It seems that before a following *w
the fricative *z failed to undergo retraction and was fronted instead
(by dissimilation?); the resulting *Dw- gave /d-/ upon the loss of *w in
clusters. The fact that we don't find /D-/ in random variation with /d-/
in this lexical set is probably significant: *dw- and *dHw- are also
consistently reflected as /d-/ in Albanian. Of course *g^(H)w- didn't
follow the same path because it became monophonemic *3'W- (> *3^ > z)
already in Early Proto-Albanian.

I don't know what happened to *sw- in unstressed syllables. I would
expect *sw- > *Tw- > *T-, but I can't offer any examples.

In <vjehërr> 'father-in-law' (PIE *swek^uros) the initial /v/ is due to
old metathesis (*swek'ur- > *wesk(')ur-), not to a special treatment of
*sw-, while <gjasht> 'six' derives from the PIE variant *sek^s (plus
secondary *-t-a:), not from *swek^s.

It seems that the combinations *k^s, *k(W)s and *sk(^) all yield
Albanian <h> /x/:

*sk(^)and-nah2 > Geg hânë, Tosk hënë 'moon'
(cf. Skt. (s')candrá- 'shining; moon')
*g^n.h3-sk(^)o: > njoh 'know'
*keh1k^-sah2 > kohë 'time' (cf. Slavic *c^asU 'id.')

Whether they fell together at the same time or not, they must have
merged as something similar to or identical with *x by Roman times,
since Lat. /sk/ --> Alb. shk and Lat. /ks/ --> Alb. fsh. However, Early
Proto-Albanian loans from Doric Greek are old enough to have been
affected: Gk. kse:nos --> Alb. huaj 'stranger, foreigner'.

Piotr