--- In
cybalist@yahoogroups.com, "tgpedersen" <tgpedersen@...> wrote:
>
>
> The development of ON -a in auslaut is
>
> in Standard Swedish -> -a
> in Standard Danish -> -& (written -e)
> in the Swedish dialects of Östergötland, Småland and Gotland -> -&
> in the Danish dialects of Bornholm and (formerly) Scania -> -a
>
> which looks confusing, if you look at dialects as representing the
> original, unspoilt language: along the former Danish/Swedish border
> the reflexes of ON -a is the reverse of those of the standard
> languages. Why is that?
>
> I think the explanation is that 'dialect' is defined by opposition
> to the standard language. Thus when recorded, the 'true' dialect forms
> are those that are the most different from the standard language. The
> actual situation is probably that on both sides both -a and -& are
> used, depending on the individual.
>
> Perhaps this is a point to be remembered, when considering dialect
> material.
>
Craig Melchert concludes 'PIE velars in Luvian' thus:
"
The evidence presented above for PIE *k^ > Luv. z may at present be
interpreted in two ways. First, in the absence of any(!) examples of
PIE *k^ > Luv. k, one could assume an unconditioned change PIE *k^ >
Luv. z (> Lyc. s?), *k > Luv. k, *kW > Luv. ku (/kW/ or /kw/). I
stress again that there is evidence for all three of these changes
before front vowels: *k^eyo- > ziya- 'lie' (> Lyc. siye- ?) etc.,
*kes- > kis^a:(i)- 'comb', *kWi- > kui-'who, which'. I readily
concede, however, that the assumption that in all of Indo-European
only the Luvian languages preserved unconditionally such a three-way
distinction is inherently suspect.
For those who find such an idea unacceptable, present evidence permits
an alternative. Leaving aside the derivation of Luv. -iza- from
*-i(s)k^o-, which is very speculative, all other examples of Luv. z <
PIE *k^ involve paradigms where a front vowel, *y or *u/w is present
in all or some forms: za/zi- 'this' < *k^o-/k^i-, ziya- 'lie' <
*k^eyo-, za:r(-za) 'heart' < *k^e:r, wazi(ya)- 'demand, ask for' <
*wek^-ye-, -za- (iter.) < *-sk^e/o-, plus azu(wa)-, zuwana/i- and
zurni. One could therefore suppose a conditioned split of PIE *k^:
*k^ > Luv. z before front vowel, *y, and *u/w, *k^ > Luv. k elsewhere.
The consistent z in the attested paradigms above would be due to
leveling, while the absence of Luv. k < PIE *k^ would be due to
chance. If one makes the ad hoc but not refutable assumption that
*k^m.tóm becomes Lyc. sn:ta via a stage *k^inta/ k^enta (or even
*k^unta), the Lycian examples of *k^ > s are also compatible with the
above conditioning. Including *u/w as a palatalizing environment is
not a problem: cf. the Armenian treatment of velars and labiovelars as
palatals after *u (e.g. dustr 'daughter' < *dhugh2ter-) and see the
remarks of Solta, IF 70 (1965) 278 ff. I stress again, however, that
even by this second scenario Luvian shows a three-way contrast of
velars before front vowels.
I will not attempt to pursue here the implications of the Luvian facts
for the vexing problem of centum/satem and two vs. three sets of
velars in PIE. It is obvious that by the strict tenets of the
comparative method Luvian requires reconstructing three sets of velars
for PIE, supporting evidence from Albanian and Armenian for such a
distinction (see for Albanian Pedersen, KZ 36 (1900) 277 ff., and
Jokl, Mél. Pedersen (1937) 127 ff.; for Armenian Pisani, Ric. Ling. 1
(1950) 165 ff.). However, more than one scholar has suggested that
this is not the only (or necessarily the right) way to view the
problem: see among many Allen, TPS (1978) 87 ff., and Shields, KZ 95
(1981) 203 ff.
"
Perhaps I should pursue it, then. If PIE *k^ was a phoneme with
allophones *k/*z (-> *k/*s), PIE *kW might have had allophones
*kW/*k (cf Latin -> Italian/Spanish) and *k occurred in loans.
Interestingly the unpalatalised velars of the kentum words of
Bangani all occur before back vowel or consonant:
http://www-personal.umich.edu/~pehook/bangani.abbi2.html
except
30 gimO 'winter' (< PIE ghi-mo-)
and the related
31 (28) gimia:LO 'the coldness which prevails before snowfall'
Now since the two other layers of Bangani are both IndoAryan one
might assume that the language contains the corresponding versions
of the list of centum words but with 'proper' satem palatalisation
(to sibilants), cf the situation in the Scandinavian languages
discussed above. The 'winter' word causes trouble here, why the
'hard' pronunciaton survived here, I don't know. But the numbers
are comforting.
Torsten