>> Yes, it's true that your idea almost seems to solve a problem.
>> However, [...]
Jens:
> Dissimilation is not unmotivated;
I wasn't done with the idea yet. I had to go for coffee. Coffee
helps me do things like... walk. God bless coffee, god bless
the underpaid coffee bean farmers too.
What I wanted to say in full is this and it's similar to
what I've been saying before but with new facets being explored.
HYPOTHESIS:
------------
You say that all disappearances of *s are really disappearances
of an entirely new phoneme *z that otherwise doesn't show its
face in IE. Yes, there are immediate credibility problems with
this basic premise of your theory that would prevent many from
just accepting it at face value. Yet there are problems here
with all our thoughts so far on Nominative Loss, including even
my own to a degree. Time to evolve a new solution.
Dealing with what we know, we see nominative *-s isn't in stems
ending in *n, *r, *l and *y when we'd expect it to be there.
This loss seems to affect only the nominative which seems to
suggest that it is somehow special from the other morphemes.
Its specialness doesn't necessarily mean that it is a different
phoneme. It could even be because of a morphological specialness
or some other unique feature that sets the nominative apart from
other suffixes. However, because of this phonetic-related loss
after certain phonemes, the specialness has to be at least a
phonetic one of some kind.
Now, at this point of inquiry, we cannot just jump to the
conclusion that phonetic means phonemic and that there is a
distinct sound *z yet. We need a very clear reason for doing
so, otherwise it's just a phonetic allophony of an already
reconstructed phoneme in IE.
Acc.pl *-ns seems to contradict this, but then I realize that
if it were still *-ms up to the time of Nominative Loss, then
*-s wouldn't disappear after *m. With *kWis, if we think of
*i as a vowel rather than a syllabized consonant, it doesn't
violate the rule either and *-s is preserved as we find.
There are then no contradictions of *-s that should be lost
or *-s that is unexpectedly preserved. Any contradictions of
phonetics involve the lack of loss of other *s-final phonemes.
So let's stick to the nominative and ignore the question of
why the genitive in *-os doesn't behave this way for now.
One problem at a time.
While the set of {*r, *l, *n, *y} seems to form a natural
class of continuants, *m is not included here meaning that
the presence of continuants before *-s is an insufficient
solution. Miguel mentions voicing alone but this doesn't
work either because consonants aren't included here, some
of which happen to be voiced.
Instead, to account for the phonetics and the above set, we
need an articulatory system to explain the loss of *-s.
Presumably, the unique articulatory quality or qualities of
the above set caused an allophony of *-s and it was this
allophone that disappeared while the unaffected variant
continued on as *s.
The simplest non-contradicting system I can think of for
this purpose is:
dental or alveolar continuants {*r, *l, *n}
palatals {*y}
The question is what region of the mouth are these
continuants pronounced in. For this phonetic solution to
work, *s cannot be pronounced with the same articulation
as the two sets above. Plus, the two sets must cause the
same effect on *s, converting it to an allophone that is
prone to disappearing.
Perhaps it could be more logical to expect *s to be an
apicodental while *r, *l and *n are alveolar. Thus, this
set would alveolarize regular *s to [S]. This in fact
can work because palatalization can likewise cause [S].
The two sets then cause the same effect as I had
discovered recently.
Voicing thus cannot be the primary factor here although
there's nothing saying that [Z] can't arise as well.
EXAMINATION:
-------------
Now we go over this new theory with a fine-toothed comb
to make sure it's not deficient.
So if alveoralization is the key then, why don't we see
the loss of *s in genitive *-os? Quite simply because it
is preceded by a vowel and none of the above continuants.
Well... then why doesn't it occur in anything but final
position? Because *s must occur after the above sets of
phonemes at the end of the same syllable as the preceding
phoneme. So we need something like *CVrs.CVC to achieve
the disappearance. The problem is that *CVrs.CVC can just
as well be realized as *CVr.sCVC syllabically. In other
words, there should be a fat chance in hell that *s can
disappear medially. It therefore makes sense why we
only observe this in final position.
In fact, back to the "specialness" of the nominative that
we are forced to conclude from the existence of Nominative
Loss, we can look at this from a different angle. The
nominative is in fact unique, syllabically speaking that is.
It is one of the few suffixes without syllabicity, being
only a consonant. This then differs from genitive *-os or
plural *-es but is interestingly similar to aorist *-s-
in that sense, which also btw produced Szemerenyi Lengthening
as we all note, yet didn't disappear like the nominative
for the reasons mentioned above.
CONCLUSION:
------------
Therefore, a new phoneme is not required at all. The solution
is thus more optimal than the *z-theory. It can be described
purely as a case of allophony of *-s. The gain with this
solution is that it doesn't conflict with the obvious
idea that nominative *-s is related to *so just as
inanimate *-d is related to the stem *to-, rather than
being divorced from this solution and being given more
laborious questions of where a nominative *-z comes from.
Further, we see how your addition of new instances of
disappearing *s where there is no need for them in the
first place is a silly diversion. There was never *-s
after *so or genitive *-syo. The *-s doesn't solve
anything because there is no problem here to solve.
You're trying to solve why *s disappeared in cases where
THERE IS NO *s PRESENT IN THE FIRST PLACE! So throw
dissimilation out the window. It's only dragging your
*z-theory down.
Rather *so never had an *s and doesn't need to because
the morphological evidence shows that *so was uninflected
for case... This is what we find. We never see a case
ending after the stem *so- so it's not like I pulled
a rabbit out of a hat like placing *z's in words where
they don't belong.
Genitive *-syo is a composite of genitive *-s after thematic
and an endingless locative *ya (later *yo- with a new locative)
with the intended meaning of *[X-s-ya Y] as "Y with which (is)
X" or simply "X's Y". So postposed *yo forms the start of a
clause "with which". To add nominative *-s (or even **z, now
senseless at this point) after *yo only exascerbates the
semantics.
A nominative *yos would convey "Y which (is) X" falsely
expressing an equation rather than the proper commitative
relationship that a locative expresses. The whole reason for
the application of *-yo here was to disambiguate it from the
identical nominative, so *s in *-syo cannot be given any
meaning other than _nominative_ in these constructions despite
the fact that we linguists know that it was originally a
genitive. Further, possession with the locative is supported
by Turkish and therefore a fully natural solution.
Your only big objection to my solution of *-syo is that you
can't imagine a stage of IE where there is no case agreement.
However, in the above construction of the thematic genitive,
there is nothing saying that there is a modified-modifier
relationship to begin with! Two nouns can have two different
cases if they need to convey something properly, even in the
grammar of IE proper.
So there's really nothing to object to with my solution but
there's everything to object to when applying *z's that
aren't there in Reconstructed IE to solve problems that
don't even exist. Shakespeare said it best: "Much ado about
nothing."
= gLeN