I was reading a recent paper by H. Craig Melchert
(
http://www.unc.edu/~melchert/anttila2.pdf) about the
genitive / possessive adjective in Anatolian. In it,
Melchert revises some of his opinions as expressed in the
book "Anatolian Historical Phonology".
The presentation of the synchronic facts in Melchert's
article is helpful. If I limit myself to the Gen.sg. forms,
the facts as presented are:
Hittite: -as (adj. -ass{a/i}-, but not in genitive function)
Palaic: -as, -as{a/i}-
CLuvian: -ass{a/i}-
HLuwian: -as (i-stems: -is), -ass{a/i}-. There is possibly
also an indeclinable form -asi(i), and another adjectival
genitive -i(ya)-. It is undecidable wheter the genitive
(written -Ca-sa) always represents -as, or that it sometimes
stands for -asa (presumably from *-oso).
Lycian: -0, -Vh, -Vhe, -Vhñ (acc.), -Vh{e/i}-, -i(je)-
Milyan: -Vse, -Vzñ (acc.), -Vs{e/i}-
Lydian: -Vl(i)-
Carian: -s', -s'ñ (acc.), -s (dative, sometimes genitive).
Pisidian: -s
Sidetic: -z
In the diachronic part, Melchert concludes that the
following PIE prototypes underlie the Anatolian forms:
PIE *-os -> Hitt. -as, Pal. -as, HLuv. -as, Lyc. -0
PIE *-oso (thematic) -> Lyc. -Vh [with syncope], -Vhe,
secondarily declined the accusatives in -Vhñ; Carian -s;
secondarly declined Pal. -as{a/i}-. Perhaps also HLuv.
-as(a), -is(a), if the -a is real.
PIE *-osyo (thematic) -> perhaps HLuv. -asi(i) and Carian
-s' (secondarily declined -s'ñ).
PIE *-eh2sos -> Hitt./HLuv./CLuv. -ass{a/i}-, Lyc.
-Vh{e/i}-.
PIE *-iyos, *-olos for the other adjectival forms.
The main weakness of this account is that it arbitrarily
separates the Hittite/Luwian possessive adjective -ass{a/i}-
from the Palaic possessive adjective -as{a/i}-. That cannot
be correct. I also do not understand Melchert's argument
for reconstructing *-ah2- instead of *-o- in the possessive
adjective: "...the first vowel of the suffix is inherently
/-a-/, and [..] all cases of /-e-/ may be due to the Lycian
umlaut rule whereby /a/ becomes /e/ before a front vowel in
a following syllable." The Lycian umlaut rule also
specifies that /e/ (< PIE */o/) becomes /a/ before a back
vowel in a following syllable. I see no objections against
deriving the Anatolian possessive adjective from PIE *-osyo-
[N. *-osyos, A. *-osyom, etc.] (> -assa-/-assi- in Hittite
and Luwian, -asa-/-asi- in Palaic, -{a/e}h{e/i}- in Lycian,
and -s' in Carian). The accusative forms Lyc. -Vhñ, Milyan
-Vzñ, Carian -s'ñ, would then be (as per Melchert's earlier
account) syncopated variants (< *-osyom).
If we take a step back from Anatolian and look at the
greater picture now, we see that the Indo-European genitive
singular has the following distribution:
1) In the personal pronouns, the genitive is related to the
accusative, the difference being one of accent (*m(m)é ~
*éme/*méne/*méme; *twé ~ *téwe; *swé, *séwe).
2) In the other pronouns, the endings are *-esyo and *-osyo
(the latter is also the thematic ending).
3) In ordinary nouns, the ending is *-os, *-es, *-s.
It stands to reason that the situation of the personal
pronouns is the oldest one. As far as I'm concerned, this
is confirmed by the gen. pl., where common nouns have the
form *-om (< **-âbhi, the word-final, proterodynamic form of
the plural oblique which appears as *-bhy- in the other
oblique plural forms), and o-stems have *-o::m < *-oy-m,
i.e. the oblique *-oy (< *-a-ati) + *-m from the athematic,
to distinguish it from *-oy used as an acc. pl. (*-oy +
[athematic] -ms > *-o:ns) and even as a nom.pl. (*-oy).
The next question is: why do demonstratives/interrogatives
and thematics have a gen. sg. separate from common nouns,
and is there a way the two (sigmatic) forms can be related?
My solution is that both the athematic gen.sg. *-(V)s and
the thematic/pronominal gen.sg. *-Vsyo are adjectival and
have the same origin. Simply said: athematic words used an
athematic adjectival declension, while thematic words used a
thematic adjective. This adjectival suffix was **-si-.
When added to the athematic oblique in **-a, the result was
**-asi, which regularly became *-os-, *-es- or *-s-,
depending on the stress-paradigm. Since it was an
adjective, it must originally have been declined, and we
would expect forms like:
G+N: *-a(:)si-z => -oss > -os
G+A: *-a(:)si-m => -osm
G+n: *-a(:)si-0 => -os
G+L: *-así-a(i) => -sí(:),
etc.
It would be nice if some of the Anatolian forms discussed
above (-hñ, -zñ, -s'ñ) could be derived from athematic
*-os-m, but I don't think that's likely (they rather come
from thematic *-osyom). The only surviving form appears to
be a fixed, undeclinable form, based on the Nm/NAn *-os
(with its variants *-es, *-s).
The expected thematic forms are *-osyos, *-osyom, etc.
Because the thematic adjectival ending *-os was added when
final *-sy was in the process of losing its palatalization,
we also find variants *-oso-, etc. I think they survive in
the Anatolian sigmatic possessive adjective, which is
therefore no innovation, but an archaism (the spread of this
form to all nouns, including athematics, in Luwian, Lycian,
Carian *is* an innovation). The use of similar adjectival
formations based on *-(e/o)lo- (Lydian, Hittite genitive of
the personal pronouns) and *-iyo- (Luwian, Lycian) shows
that this was no isolated phenomenon.
As with the athematic genitive, the thematic possessive
adjective also had a tendency to become undeclinable, and
this fixed suffix appears as *-osyo, both in Anatolian (Lyc.
-Vhe, perhaps Carian -s' and HLuw. -asi(i) [unless that's
merely a spelling variant of <-a-sa>, i.e. /-as/], as well
as in the rest of IE (as *-esyo ~ *-eso, *-osyo ~ *-oso).
As Melchert says, the possessive adjective does not normally
allow expression of the plurality of the possessor, but
CLuvian developed a variant -anzassa/i- to mark precisely
that. Conceivably, the pronominal gen.pl. *-oyso:m is a
similar device (based on obl.pl. *-oy and the gen.pl. ending
of the poss.adj. *-osyo:m).
=======================
Miguel Carrasquer Vidal
mcv@...