I said I was going to read Françoise Bader's "Problématique
du génitif thématique sigmatique", and I have, but I've been
struggling with what to say about it. I disagree on almost
everything...
The article starts by saying (under SOMMAIRE): "Le génitif
thématique ancien a été en *-os, homonyme du nominatif". I
don't think Bader has quite succeeded in proving that. That
the thematic (and pronominal) genitive was once simply *-os
is a theory that has certain things in its favour, and I
don't wish to reject the posibility (see below), but Bader's
arguments in favour of its retention in PIE are weak. The
best one is of course Hittite thematic -as. Leaving aside
preconceptions, there is a good chance that this continues an
original them.gen. ending *-os, and there is an equally good
chance that it represents analogical spread of athematic *-os
to the o-stems. I don't think Old Prussian -as (G. <deiwas>)
offers much support (the nom. is <deiwas> or <deiws>, which
makes it far more likely that G. <deiwas> (never <deiws>) is
/deiwa:s/, i.e. the general Balto-Slavic thematic ablative
turned genitive *-aa (> *-a:) supplemented with analogical
-s). Celtiberian -o (not an instrumental, which is -u, nor
an ablative, which is -uz) in my view represents *-os(y)o >
*-o(y)o > *-o:, and is thus fully comparable to Greek -ou (=
/-o:/) < *-o(y)o < *-os(y)o. The explanation quoted by Bader
(sandhi variant of *-os) is absurd. Bader's attempts to see
*-os in Vedic rathas-pati- (why is German Rad an s-stem?) and
Greek theós-dotos, Lukós-oura (analogical after Dios-dotos
and Kunós-oura, with athematic genitives), and to suspect a
reading /-os/ in Homeric verses where the text has "Attic"
-ou (instead of the more common -oi[o]) are not very
convincing.
Whether a putative thematic genitive *-os would have been
originally homonymic with the nominative is doubtful. The
nominative goes back to *-oz, while the genitive is properly
*-es, which would give *-o-(e)s > *-os in its thematic form.
Homonymy in this scenario would have been late (after *-z >
*-s and zero-grade), but may indeed have been a motivation to
modify the them.gen. form, as Bader states.
In Bader's view, the old form *-os was replaced by other case
forms, such as the i-stem genitive -eis in Osco-Umbrian, or
the ablative in Balto-Slavic. Or by the relational suffix
*-ih2, as in Celtic and Latin. Or by the adjectival suffix
*-syo- (or its doublet *-so).
I agree that Balto-Slavic replaced the thematic genitive
(*-osyo) by the ablative. I go further than Bader in this,
by claiming that the pronominal genitive in Slavic (-ogo,
-ego) is also an old ablative (*-ood, *-eod) [Bader, like
most people, misguidedly sees here an "emphatic" element
-go]. I'm not so sure about Osco-Umbrian -eis. Bader
mentions some old Campanian inscriptions (ieviesiie sum,
brutesiie sum), which may represent old genitives in *-osyo
(> -esiie). In view of the rest of Italic and Celtic, *-osyo
(and *-esyo) must have existed in Osco-Umbrian as well, and
may have as been prone to lose the /s/ as elsewhere in
Italo-Celtic. I'm not sure if unstressed -oi- can give -ei-
in Oscan (loc. sg. *-oi gives -úí), but *-esyo > *-eyo > *-ei
is I believe wholly justified (for loss of final short
vowels, cf. *hortos > húrz). Addition of genitive -s would
give attested -eis (> -es in Umbrian). A similar form -es
(-e:s?) is also known from East Celtic (Pannonia, etc.).
Bader does not even mention the possibility that Latin (where
-osio is attested in the oldest inscriptions) may continue
*-esyo and *-osyo in the pronouns (*esyo > eiiu+s, eiu+s,
*kWosyo > quoio+s, cuiu+s, *ghosyo(?) > huiu+s, unstressed
-i:u+s in isti:us, illi:us) and, by a reduction *-i:u > -i:,
also in the o-stem genitive [the thesis advocated by
Szemerényi]. The same goes for Gaulish and Insular Celtic,
where *-osyo, *-yosyo and pronominal *-esyo, reduced to
*-oyo, *-yoyo, *-eyo, may well be the origin of genitive -i:
(the nom.pl. *-oy > -i: is attested still as -oi in an
inscription together with gen. -i:, so the most liklely
hypothesis is that -i: originates in soft-stem *-yoyo or
pronominal *-eyo [cf. Gothic nominal -es < *-esyo]).
The other Celtic forms (Lepontic [and Venetic] -oiso <
*-osyo, Celtiberian -o(:) < *-oo < *-os(y)o) are also
favourable to *-osyo (*-esyo) as the original them. gen. in
Celtic.
Bader's treatment of Baltic and Germanic I don't comprehend.
The gen. form of the Old Prussian pronoun stas (stesse etc.)
is taken at face value as containing a geminate /ss/, which
in reality only reflects German orthography, and this is then
equated to the Germanic genitives -es/-is and -as/-æs,
reconstructed as *-esso, *-osso, because of the lack of
rhotacism in these forms. What about Verner? Clearly all
these forms reflect *-es(y)o and *-os(y)o, without any
problems.
I'll skip Bader's reflections on the origin of the pronominal
declension (they start with: "La flexion pronominale n'a rien
à voir avec la flexion nominale").
I see no reason to review my thoughts on the Anatolian
genitives and genitive-like constructions. Luwian
-a(:)ss(a/i)-, the "adjectival genitive", can be equated with
Hittite -assa- and Palaic -asa- (which give relational
adjectives). Melchert's Lycian -ahe/-ahi- (but also
-ehe-/-ehi-) is inconclusive to establish Proto-Anatolian *a
rather than *o, and *o fits better with the lenghth of Luwian
-a:ss(a/i)-. Hittite and Luwian share a soundlaw which
geminates /s/ before a resonant (e.g. Melchert 10.1.6.1.1.3
p. 266), a development not shared by Palaic and presumably
Lycian (-s- > -h- in -ahe-/-ehe-), so *-sy- > -ss- in Hitt.
and Luw. is as expected, as is *-sy- > -s- in Palaic and
Milyan, *-s- > -h- in Lycian. Carian gen. -s' (with another
sibilant than plain /s/, but not /s^/ either), may stand for
a special development of *-sy- (palatal /s'/?) in that
language.
In summary, instead of the chaotic mess suggested by Bader,
we have, in my opinion:
Celtic
Lepontic: (old) -oiso, (new) -i(:)
Gaulish/Insular: -i: (< *-i:o < *-eio < *-esyo/*-yosyo)
Eastern: -e(:)s (< *-es(y)o + -s?)
Celtiberian: -o(:) < *-oo < *-oso
Italic
Latin: (old) -osio, (new) -eiu(s), -uiu(s), -i:u(s), -i:
Osco-Umbrian: (old) -esiie(?) (new) -eis, -es < *-esyo + -s
Venetic: -oiso < *-osyo
Germanic
Gothic/Nordic: -es < *-esa < *-es(y)o
West Germanic: -as, -æs (> -es) < *-asa < *-os(y)o
Baltic
Old Prussian: -a(:)s < *-a: + -s
East Baltic: -a: < *-oo < *-ood (ablative)
Slavic
-a < *-ood, pronom.: -ogo, -ego < *-oho, *-eho < *-ood,
*-eod; *-eso in the pronoun c^eso.
Greek
*-osyo > -oio (Hom.)
*-oso > -oo > -ou
Armenian
*-osyo > -oy
Indo-Iranian
Skt. -asya, Av. -ahya < *-osyo
Anatolian
Hitt./Pal./(Luw.) -as < *-os
Hitt./Luw. -assa-, -assi- < *-osyo-
Pal. -asa-, Lyc. -ahe/i- ~ -ehe/i- < *-os(y)o-
Sid., Pis. -s < *-os(y)o(-)
Carian -s' < *-osyo(-)
That is enough to warrant the reconstruction of pronominal
*-esyo/*-osyo, nominal *-osyo (and *-yosyo) for at least the
whole of non-Anatolian Indo-European.
Bader's article (after the Summary and the Plan) begins with
considering the Hitt. noun wastulas "sinner", derived from
wastul "sin", as if meaning "le du péché" (he of sin).
However, the ending -as in <wastulas> is simply the thematic
nominative (if it's a nominative) or the Hittite thematic
genitive (if it's a genitive). The thematic declension may
originally have conveyed something like a definite meaning,
but one cannot attribute the double function of definite
marker _and_ genitive marker to the single morpheme *-os.
Forgetting about <wastulas>, the thematic genitive _is_ a
place where we would expect a double marking for
"definiteness" (or simply "thematic-hood") and genitivity to
co-occur. Perhaps then the link between general PIE
*-os(y)os and the Luwian/Anatolian relational adjective
*-osyos (*-osyod, *-osyoi, etc.) can be established by
positing forboth an origin in a doubly marked *-os-os (or,
more accurately, *-os^ + -oz), i.e. the genitive (*-os^)
followed by a thematic ending (e.g. nominative *-oz).
The distinction between *-osyos and *-osos is generally
disregarded, but Bader maintains the two strictly separate
(although stating at the beginning that *-so is a "doublet"
of *-syo). So what _does_ explain the alternation (often
within dialects of the same language e.g. Greek -oio and
-oo)? I see a parallel with the plural oblique morpheme
*-bhi-, alternating with *-bh- and *-m- (e.g. Skt. -bhyas,
Lat. *-bos, Slav. *-mos). When it stands in the absolute
Auslaut (namely in the G.pl. *-om), only the shape *-m- is
found, but when it has been built upon to make additional
oblique case endings, all three variants occur.
Just like the oblique plural suffix comes from earlier
*-(a)bhi, the (athematic) genitive suffix might come from
earlier *-(a)si (cf. Etruscan gen. *-si). In the Auslaut,
this regularly gives *-os/*-es/*-s (depending on the accent
paradigm). But when thematic *-o- has been added, we see the
variants *-o-sy-o and *-o-s-o, as if the thematic vowel had
been added at a time when final *-i was just in the process
of being lost.
The co-occurrence in Hitt. of thematic -as and a class of
relational adjectives in -assa- perhaps *does* vindicate
Bader's thesis that Hitt. -as continues the old thematic
genitive in *-os. In that case, the first stage of
development (seen in Luwian) was the replacement of *-os (and
pronominal *-es/*-os) by the relational adjectives *-osyos,
*-osyod, *-osyah2 etc. The second stage (seen in part in
Lycian, as well as in Pisidian, Sidetic, Carian and in
non-Anatolian IE) was the deflexivization of the genitival
adjective, turning it into a genitive proper. In Pisidian,
Sidetic and Carian this was semi-automatic by the loss of
final vowels, in "non-Anatolian PIE", this was accomplished
by generalizing *-osyo (perhaps dissimilated from *-osyos),
or its variants *-oso, *-esyo, *-eso, and fixing these forms
in the genitive spot of the pronominal and o-stem paradigms.
For the rest, as we have seen, *-es(y)o and *-os(y)o are
reconstructible for the vast majority of Indo-European
languages.
=======================
Miguel Carrasquer Vidal
mcv@...