OK. I'll try one more time.
Suppose there was a phonetic split that caused an adjectival suffix,
whatever its original shape, to split into two.
After the split, we have two _separate_ suffixes: *-u and *-ro.
_We_ know they are historically related, because we have the comparative
data, and we know Sanskrit and Latin and stuff, but if you're a peasant
tilling the land on the banks of the Nemunas or the Tarim, I don't think
that by the sound of it you'd know they were related.
So how did the pre-Tocharians know, when they decided that adjectives
should be thematic, that they were supposed to replace those cumbersome
u-stems with ro-adjectives, instead of, say, -wo, or -no, -mo, -to, -ko, or
plain -o? And how did the Lithuanians, when they decided that it was
boring to have so many thematic adjectives, know that they were supposed to
replace ro-adjectives with u-stem adjectives, and not i-stems, or r-stems,
or whatever?
That sounds as crazy as suggesting that pre-proto-Germanic took part in de
satem shift, but after a while the proto-Germans decided to reverse that.
Basically, there seem to me to be four possibilities.
1) Crazy as sounds, this is what happened. After all, speakers of Egyptian
Arabic indeed did revert the ji:m-phoneme to its pre-Classical (and
proto-Semitic) pronunciation /g/. The fact that there was _no merger_
involved makes it at least possible, even if unlikely.
2) The whole thing (sonant/vocoid soundlaw) is an artefact of the surviving
data. That's the "alternative scenario" I suggested.
3) At the time of the breakup of PIE, *-u and *-ro shared something
_semantical_, which was not shared by any other adjectival suffixes.
Whatever it was, I cannot recognize it in the semantics of the attested u-
and ro-adjectives. They are just adjectives, and they don't seem to share
a common semantic overtone, different from that of i-stem or that of other
thematic adjectives.
4) At the time of the breakup of PIE, *-u and *-ro shared something
_phonological_, which was not shared by any other adjectival suffixes.
Whatever it was, it has since been lost, given the fact that /u/ and /ro/
do not share much phonetically. However, I'd like to stress here again
that Armenian u-stem adjectives have a nom/acc. sg. in -r. Since Armenian
derives from PIE, it's not totally unreasonable to derive that -r from *-ur
in PIE, if only for the NA sg. _neuter_ of u-stem adjectives.
=======================
Miguel Carrasquer Vidal
mcv@...