> Ämne: Re: [CONLANG] Mismatched phonologies / accents Från:
> Peter Collier <[log in to unmask]> Datum: Fri, 25
> Jan 2008 01:03:32 -0000 Till: [log in to unmask]
> ----- Original Message ----- From: "Benct Philip Jonsson"
> <[log in to unmask]> To: <[log in to unmask]> Sent:
> Thursday, January 24, 2008 4:37 PM Subject: Re: Mismatched
> phonologies / accents
> ...
>> If the language having the inventory is some kind of pre-
>> Old High German I'd expect /s/ to be apical [s_a] (like
>> in modern Spanish and Greek), and /z/ not to exist[^1]
>> and thus:
>> /ts/ > /T/ /dz/ > /D/ /S/ > /x/ /Z/ > /j/ or /G/ /tS/ >
>> /S/ > /x/ or > /tj/ > /tt/ (/t/ word initially) /dZ/ >
>> /Z/ > /j/ or /G/ or /dj/ > /dd/d/
> Well spotted, the inventory is PGmc. The /s/ is [s_a], so
> I suppose to be ultra precise I should list WRom [s_m] >
> NRom [s_a].

That's not a mere nitpick, since [s_a] is more [S]-like and
[s_m] is more [T]-like. In fact in modern Icelandic "/T/"
is [s_m] and "/s/" is [s_a], both alveolar. It is known
that in the history of Spanish the following series of
changes took place:

| s > s_a
|  > z_a > s_a
| k', kj, tj > ts_m > s_m > T
| k' > dz_m > z_m > s_m > T
| g', j > dZ > S > x
| S > x

and it is believed that Old French had a similar situation,
although there /s_a/ and /s_m/ merged with each other
instead, and likewise with /z_a/ and /z_m/.

So we can be quite assured that at some time OTL Western
Romance had the following sibilant system:

| ts_m   s_a  (tS)   S
| dz_m   z_a   dZ   (Z)

where the parenthesized items were either rare or lacking in
some areas.

Comparing this to the pre-West Germanic fricatives system
I'd not be the least surprised if Germans learning Gallo-
Romance would equate the foreign [ts_m] with their [T],
especially if there wasn't yet any /ts/ in their Germanic
language. The biggest problem to me is what they'd make of
[S]. There was perhaps no x to equate it with any more,
since Old High German consistently keeps /h/ from Germanic
*x and /x/ from Germanic *k distinct, in which case I'd have
[S] merge with /s_a/ in substratization. OTOH with a time of
contact as early as the first century I'd slate [S] to be
equated with *x and then develop to a /h/ distinct from the
lost Latin *h.

I have wondered for long how k' merged with tj in Gallo- and
Ibero-Romance and come to the conclusion that it was *not*
via a progression k; > c > tS > ts but rather that k' merged
with tj directly. It is believable in that at least to the
naked ear the two palatalized sounds k' and t' sound much
more similar to each other than the non-palatalized k and t.
But how come then that when g' and d' merge they both become
/dZ/ and not both /dz/? Perhaps simply because d' was so
infrequent to begin with? The variant spellings DAZA and
DAIA for the same name suggest that merger in some direction
did happen. An older generation of Romanists thought that g'
in all positions went through a [j] stage before becoming
/dZ/, and that this together with a prestige pronunciation
of the letter _z_ as d' worked against a merger. It is
notable that the lenited reflex of k' is _dz_ but that of g'
is [j]!

(There is no "alveolar" diacritic in CXS: [tdnszl] are
alveolar by default. I guess the 'retracted' [_-] diacritic
could be used as well (which would then agree with the way
Dravidianists mark alveolar consonants with a subscript
minus); [d_+] could then mean a "postdental" [d] as opposed
to interdental [d_d] if one wants to.)

> /z/ of course had earlier rhoticised to /r/, I'd forgotton
> about that!

My forgotten note was to be that the phoneme which Keller
writes as /z/ is actually /s_m/, traditionally transcribed
by Germanists with U+0225 LATIN SMALL LETTER Z WITH HOOK. I
consider that a nasty feature in that book.

And of course you are right, Gmc. *z > r in WGmc., except
*-z which > 0.

> In which case then I guess I'd also need /z/ > /s/ [z]. Do
> you think [z] would ultimately devoice and merge with
> [s_a], or would it survive long enough to become a
> separarte phoneme again?

I thimk it would merge with [s_a]

> I'd also completely fogotten about /L/ My guess is
> that would go to /j/, although maybe /r/ is a
> possibility there?

/L/ usually becomes /lj/ when borrowed into Gmc.
languages; consider Swedish _fåtölj, portfölj_. /J/
likewise > /nj/: _konjak_.