I'm reading Eric P. Hamp's idiosyncratic article "Whose [sic] were the
Tocharians?" (in: Victor H. Mair "The Bronze Age and Early Iron Age
Peoples of Eastern central Asia"), where I find an interesting
argument from Tocharian for PIE *h3. We have: Toch. B. <n~em>, A.
<ñom> "name" < *h1nomn, versus Toch. A. <maku>, B. <mekwa>
"fingernail" < *h3noghw-. According to Hamp, *h1 caused the
palatalization of *n > ñ, while *h3 caused the rounding of *n > m.
Nothing to do with *h3, but it's nevertheless curious that I had
envisaged much the same developments for pre-PIE **n. To
recapitulate:
**-n- generally stays as *-n-.
In the Auslaut, **-n > *-r, except when the preceding consonant is
*/m/. This explains the PIE -r/-n- heteroclitics (all neuters, so
with *-n in the absolute Auslaut in the nom/acc). It also explains
the neuters in *-m(e)n, which failed to undergo the shift. Also
explained is the 3pl. preterite *-e(:)r, which can be analyzed as zero
ending + plural morpheme *-en- (**-an-), to which additional 3rd.
person *-t (or *-s, e.g. in Indo-Iranian) could optionally be added.
The symmetrical scheme is:
*-mw > *-m *-mw-án > *-mén/*-wén (*-me(s)~*-mos, *-me:(s))
*-tw > *-s *-tw-án > *-tén (*-te(s), *-te:(s), etc.)
*-0 > *(-t) *-0-án > *-ér(-t) [present *-enti]
The obvious exception is the 2pl., where we would expect *-ter (vs.
present *-teni). Tocharian B. has -cer (< *-te:r), which may be an
isolated remnant of this expected form (with unexpected long *e:, as
in Baltic *-te:, OIr. -the), and there's also the intriguing Vedic
-thana/-tana (besides -tha/-ta). It's probably too much to expect
Albanian -ni to come from *-teni.
Other exceptions (i.e. PIE words ending in *-n) are rare: there is *en
"in", which, as a preposition, was never in absolute Auslaut, and may
be short for *en-i (with Loc. *-i) in any case. There are some
interesting locatives in *-r [< **n?] (such as E. "where") as well.
Any other PIE *-n's?
For the apparent exception *[...]newn "9", see below.
Palatalized **-n^ can give both *n and *i. This is apparent in a
number of lexical items (*nem-, *yem-, *em- "to take" < **n^em-; or
maybe *su:nus/*su:yos "son" < **suH-n^- (?)), but morphology provides
another clue in the *-i/*-n- stems (such as found in Vedic), e.g. N.
pátis, G. pátyur, fem. pátni: or N.A. ásthi, G. asthnás, where the *i
in the nom.sg. might well be derived from earlier **n^.
Finally, for labialized **-nw > *-n- / *-m ~ *-w-, we can adduce the
numeral "9", which we know had *n from the ordinal (Lat. <no:nus>),
and is therefore usually reconstructed as *-newn., although the
reflexes may just as well be explained by *-newm., which is in fact
the reconstruction I would suggests as the correct one. It's possible
that the *-/m./ was analogically transferred from *septm. and *dek^m.
(where it appears to be etymological), but another possibility is that
it was the lautgesetzlich continuation of pre-PIE **náwnw (<
**naunu/**nu:nu).
Whether the word for "9" has anything to do with "new" or not, there
is the possibility of *neu-i-os < **nawn^-os and/or Arm. nor "new" <
*nowor < *nowr. < *na:wn.
If we put all of this together and combine it with the possibility of
thematized fossil nominatives/absolutives, we get something that
approximates the Caland system very closely. Suppose we had a
derivative suffix **-Vn-, with endings **-nu, **-na and **-ni (nom. ~
erg., acc. ~ abs., gen. ?), then we might expect an old genitive to be
maintained in compounds (*bhrg^h-i- < **bh(e)rg^h-n^-), the "active"
form **bh(e)rg^h-nw to have yielded *bhrg^h-ú- or thematized
*bhrg^h-m-ós, while the "passive" form **bh(e)rg^h-n(-) might appear
as such, or as thematized *bhrg^h-r-ós. It would be interesting to
see if there are indeed traces of an active/passive or
transitive/intransitive bias in the PIE adj. formations in *-u/*-mós
vs. *-rós/*-n-(os). (*-mós, the verbal adjective, seems to be both
active and passive, while *-nós is mainly passive).
=======================
Miguel Carrasquer Vidal
mcv@...