Richard:
> Doesn't the proposed analogical change in the thematic 2s
> *-&z > *-&s following *-&si thereby imply a phonemic contrast
> between final /s/ and /z/?
Actually, no, it doesn't need to. If we have [-&z] and [-&si],
there is no contrast in the final position at this stage and
[z] remains an allophone. We only need to propose levelling
of the thematic vowel to *e based on the default indicative
where voicing never existed.
Eventually, although not necessarily, [-z] can have changed back
to [-s]. I say not necessarily because we might still propose [-z]
in the latest stage of IE itself. In this way, it would still serve
to explain the Nominative Loss in the same way as Jens does without
needing a new phoneme. We only need to use an allophone of *s
which we already know exists in *nisdos [nizdos/nizdoz].
In other words, this theory gains by reusing existing elements
that IEists already accept and are familiar with (ie: the z-allophone
of *s) rather than going through the laborious process of trying to
substantiate new phonemes whose proof remains far more obscure.
> If one's willing to accept analogical explanations though, what
> stops secondary *-os being remodelled as *-es after primary
> *-esi? This seems the simplest explanation if one won't accept a
> contrast /s/ ~ /z/.
Well, there are different places where the remodelling can occur.
The differentiation of the primary and secondary 2ps and 3ps
endings is trivial at that stage if we accept the allophony
solution. After that, Schwa Diffusion occurs and then Vowel Shift.
You are proposing something occuring after Vowel Shift. Let's look
at this phonetically stage by stage.
These are the results of what you're saying:
INDICATIVE NON-INDICATIVE
[-&-si] [-&-s]
Final Voicing [-&-si] [-&-z]
Schwa Diffusion [-&-si] [-&.-z]
Schwa Merger [-e-si] [-a-z]
Vowel Shift [-e-si] [-o-z]
2/3ps levelling [-e-si] => [-e-z] (!!!)
*-e-si *-e-s
The problem I have with this is that the levelling can occur more
credibly during the stage during Schwa Merger where half-long *&.
merges with full vowel *a and *& with *e (the latter probably the
first change to occur before *&. becomes *a). So in my view, a
less phonetically dramatic analogical levelling would be as follows.
INDICATIVE NON-INDICATIVE
[-&-si] [-&-s]
Final Voicing [-&-si] [-&-z]
Schwa Diffusion [-&-si] [-&.-z]
Schwa Merger [-e-si] => [-e-z]
Vowel Shift [-e-si] [-e-z]
2/3ps levelling [-e-si] [-e-z]
*-e-si *-e-s
As I said, *& > *e might have occurred first before *&. > *a,
meaning that there are two halves to Schwa Merger. This would
be an even more trivial solution since:
[-&-si] [-&-s]
Final Voicing [-&-si] [-&-z]
Schwa Diffusion [-&-si] [-&.-z]
Schwa Merger I [-e-si] [-&.-z]
[-e-si] => [-e-z]
( Schwa Merger II [-e-si] [-e-z] )
Vowel Shift [-e-si] [-e-z]
2/3ps levelling [-e-si] [-e-z]
*-e-si *-e-s
I think that this would be a more optimal solution because it
doesn't give us e/o-whiplash :) The unexpected change of
*&. => *e is comparatively lesser than *o => *e. Some side
conclusions of this would be that by the time of Schwa Merger II,
the thematic 2ps secondary was now identical with the plural.
The plural, as I've said to Jens, had escaped Schwa Diffusion
because it always had *e, not *&.
= gLeN