A new overview, taking into account the discussion so far.
o-stems:
PIE PGmc
Nom. *-os *-az
Runic -az, Goth. -s, ON -r, WGmc -0.
Acc. *-om *-aN
n. *-om *-aN
Runic -a, ENWGmc. -0.
I think we're all agreed that in the famous inscription "ek hlewagastiz
holtijaz horna tawido", the form horna can only be a NA sigular. There is
no way it could be a dual or a plural.
Voc. *-e *-i > -0
Gothic -0, elsewhere merged with or replaced by the nom.
Gen. *-osyo *-asa
Runic -as, OE (West-Saxon) -æs > -es, OS -as. Gothic -is comes from the
pronominal ending *-esyo > *-esa. OS/OHG -es are problematical. They
could be directly from *-esyo > *-esa > -es, like the Gothic form, but one
would prefer an explanation with *-as for the whole of N. and W. Germanic.
Some of the OS forms are apaprently early cases of reduction to schwa of
unstressed vowels, and for the rest, Jens' explanation of analogy after
pronominal -es (des dagas > des dages) seems adequate (that's *-esyo too,
but independently from Gothic).
Dat. *-o:i *-ai
Loc. *-oi *-ai
North and West Germanic -e could come from either Dat. or Loc. Gothic -a
cannot come from either (we'd expect -ai), so it's a instrumental.
Ins. *-o: *-o
Gothic -a (dat.), OS/OHG -u. OE -e is a dative/locative.
Abl. *-o:t *-o:
Not preserved as a case form, but found in adverbs (Goth -o:, ON -a, OE -a,
OS -o, OHG -o)
pl.
The nominative in *-oi is unattested in Germanic. PIE *-o:s would have
given PGmc. *-o:z, which explains Goth -o:s and ON -ar, but not OE -as, OS
-os. The reconstruction is thus:
PIE *-ó:ses *-o:siz
*'-o:ses *-o:ziz
which explains all the forms (Goth. -o:ss > -o:s, ON *-o:ziz > -arr > -ar,
OE/OS *-o:siz > *-as). OHG -a is the acc. form, or (if it's really -a:,
as
suggested by Kienle), it might be from *-o:ziz, as the ON form, although I
prefer to stick to W.Gmc unitary *-o:siz.
Acc. *-ons *-anz
The PIE form can be *-ons or *-o:ns. Germanic *-anz (Goth -ans, N/W -a)
must be from *-ons (but that can be from *-o:ns by a shortening law, such
as Osthoff's).
n.pl. PIE *-&2 > *-a or *-eh2 > *-a:. Germanic continues the second form
(Goth -a, OE/OS -u ~ -0, ON/OHG -0).
Gen. *-o:m *-o:N
Not problematical in N/W Gmc. (ON -a, OE -a, OS -o, OHG -o), where the
ending has spread to all declensions. For my latest thoughts on Gothic
-e:, see the i-stems.
The Dat/Abl. was PIE *-oios, *-oibhios or *-obhios, Ins. *-o:is. In
pre-Germanic, the Ins. was probably regularized to *-obhis > *-amiz and
merged with *-obhios > *-amjaz to *-amiz. This explains Goth. -am,
but I'm still confused about the other forms. ON and OE have -om and -um
respectively as they have in all declensions, OS and OHG have -um.
Since OHG has -ame:s in the 1pl. of thematic verbs, I find it hard to
swallow that PIE *om gives NWGmc *um. But I have no better explanation
myself. Any more details, ideas, opinions?
The Loc.pl. in *-su has not left any traces in the Germanic declensions.
In summary:
N *-az *-o:siz, *-o:ziz (, *-o:z?)
A *-aN *-anz
n *-aN *-o
V *-i
G *-asa, *-esa *-o:N
D *-ai *-amiz
I *-o
(Ab. *-o:)
a:-stems
PGmc
Nom. *-eh2 > *-a: *-o
Gothic -a, ON -0 (*-u), OE -0 ~ -u. OS/OHG -a from the accusative.
Acc. *-eh2m > *-a:m *-oN
OE -e (*-a), OS/OHG -a. Gothic and ON have the nominative form.
Voc. = Nom.
Gen. *-eh2os > *-a:s *-o:z
Goth. -o:s, ON -ar, OS -a, OHG -a:. OE -e (-ae) after the Dat.
Dat/Loc *-eh2(e)i > *-a:i *-ai
Goth -ai, OE -e. The ON, OS and OHG forms are instrumentals.
Ins. *-eh2(e)h1(?) > *-a: *-o
ON -0 (*-u), OS -u, OHG -u (all datives).
Plural:
Nom. *-eh2es > *-a:s *-o:z
Goth. -o:s, ON -ar, OE -a, OS -a, OHG -a:.
Acc. *-eh2ns > *-a:ns *-o:Nz
Merged everywhere in Germanic with the Nom.pl.
Gen. *-eh2om > *-a:m *-o:N
Goth. -o:, ON -a, OE -a, OS -o. The by-form *-eh2-n-o(:)m > *-a:no:m (Skt.
-a:na:m) is found in OE -ena (vowel /e/ from the n-stems), OS -ono, OHG
-o:no.
Dat/Abl/Ins(/Loc) *-a:-bhi(o)z > *-o:miz
Goth. -o:m, OHG -o:m.
In summary:
N *-o *-o:z
A *-oN *-o:Nz
G *-o:z *-o:N, *-o:no:N
DL *-ai *-o:miz
I *-o
The Germanic *-ja and *-jo:-stems (*io- and *ia:-stems) need not be
distinguished from the o- and a:-stems at the PGmc. stage (unless some
cases of ja:-stem Nsg. *-i: are relicts from the *ih2-stems). Otherwise, I
guess they were declined regularly, with *-j- preceding the case-endings.
The i-stems:
Nom. *-is > *-iz
Goth. -s, ON -r, OE -e, -0, OS -0, -i, OHG -0, -i
Acc. *-im > *-iN
Goth. -0, ON -0, OE -0, -e, OS -0, -i, OHG -0, -i
Voc. *-ei or *-i > *-i(:)
Goth -0 from *-i. Elsewhere = Nom.
Gen. *-eis should have given Gmc. *-i:z. Instead we find Goth. -ais,
ON -s, OE -e, OS -i, OHG -i, of which only the OS and OHG forms can be
from *-i:z. The WGmc forms are indistinguishable from the dative/locative,
so perhaps Gothic -ais is simple dative *-ai + -s (with some help from
u-stem
G. -aus, D. -au).
Dat *-eiei
Loc *-e:i *-ai
The dative forms are old locatives (Goth. -ai, OE -e, OS/OHG -i).
Ins. *-ih1 *-i:
OS/OHG have added -u from the o-stems (> -iu). OE -e is dative/locative.
Plural:
Nom. *-eies *-i:z
Goth. -eis, ON -er, OE -e, OS -i, OHG -i.
Acc. *-ins *-iNz
Goth. -ins, ON -e, OE -e, OS -i, OHG -i (merged with nominative).
Gen. *-eiom *-eiaN
ON and OE have adopted the o-stem ending *-o:N (ON -a, OE -a).
OS -io, OHG -eo may come from *-jaN, but Gothic -e: may perhaps be
explained by following van Coetsem's explanation of PGmc *e:2 from PIE *ei
with a-Umlaut.
In the G.pl., *-eio~m would have developed to *-eeo:N with *ei > *e:2
before *o:. In Gothic, this would have given *-e:a (*-e:o:?), whence -e:.
In OS/OHG the expected form would be *-e:o, whence *-eo (-io).
In ON/OE we would expect *-e:a, which perhaps survives in OE Gpl. forms
ending in -i(y)a.
Another explanation, endorsed by Jens and Piotr, is that the Gothic
ending -e: spread from the o-stems after the replacement of Gsg. *-as(a)
by pronominal *-es(a). Because *a and *o(:) formed in Germanic a
short/long pair, this would have prompted the replacement of Gpl. *-o:N by
its fronted counterpart *-e:N.
D/Ab/I/L pl. *-i-bhi(o)s > *-imiz
Goth., OHG -im.
In summary:
N *-iz *-i:z
A *-iN *-iNz
n *-i [*-(i)yo]
V *-i
G *-aiz *-eeo:N (?)
DL *-ai *-imiz
I *-i
The u-stems:
Nom. *-uz *-uz
Goth -us, ON -r, OE -u, -0, OS -u, -0, OHG -u, -0
Acc. *-um *-uN
Goth. -u, ON -0, OE -u, -0, OS -u, -0, OHG -u, -0
Voc. *-eu *-au
*-u *-u
Both forms are found in Gothic -au, -u.
Gen. *-ous *-aus
Gothic -aus, ON -ar, OE -a, OS -o, OHG -o:.
Loc. *-e:u *-iu
*-o:u *-au
Gothic -au, Runic -iu > ON -i, -e, OHG -iu from *-eu.
OE -a, OS -o (and possibly Goth. -au) from *-au
Plural:
Nom. *-eues *-iwiz (*-iwz)
*-oues *-awiz
Goth. -jus (*-iwz), ON -iR, -er, OS/OHG -i.
OE. -a from *-awiz, besides OE -u from the Acc.
Acc. *-uns *-uNz
Goth. -uns, ON -o, OE -u, OHG -u. But OE -a, OS -i, OHG -i
are nominative forms.
Gen. *-euo(:)m > *-ewo:N
Goth. *-iwo: > -iwe:. ON and OE have the o-stem Gpl (ON -a,
OE -a), OS -io, OHG -eo look like i-stem Gpl.'s, but may perhaps
also derive from *-ewo:N.
DLIAb *-ubhi(o)s > *-umiz
Goth., -um.
In summary:
N *-uz *-iwiz, *-awiz
A *-uN *-uNz
n *-u
V *-au, *-u
G *-auz *-ewo:N
DL *-iu, *-au *-umiz
The C-stems were in PIE:
strong:
N *-s (< **-z)
A *-m.
V *-0
n *-0
weak:
G *-és or *-os
D *-éi
L *-i
I *-éh1 or *-ét
Ab [wanting, expected *-ot]
strong:
Np *-es
n. *-&2
weak/strong:
A *-n.s
weak:
G *-om or *-o:m
doubly weak:
DAb *-bhiós
I *-bhís or *-bhí:s
L *-sú or *-sí
In Germanic, that should give:
Goth. ON OE OS OHG
N *-z baúrg-s mOrk [5] burg burg burg
A *-uN baúrg [1] mOrk burg burg burg
n *-0 (-0)
G *-iz (*-az) baúrg-s merk-r [6] byrg burges [7] burg
D *-i(:) baúrg mOrk byrg burg burg
N *-iz baúrg-s merk-r byrg burg-i [8] burg
A *-uNz baúrg-s [2] merk-r byrg burg-i burg
n *-a (-0)
G *-o:N (*-aN) baúrg-e: [3] mark-a byrg-a burg-o burg-o
D *-(u)miz baúrg-im [4] mOrk-om burg-um burg-un burg-um
[1] for expected baúrg-u, analogical after e.g. o-stem -s/-0.
[2] the acc.pl. has merged with the Npl. everywhere.
[3] with Gothic Gpl. -e: instead of -o:
[4] -im for -um after the i-stems?
[5] ON Nsg. -0 < Asg. -0 for expected *-r.
[6] ON/OE Umlaut shows that PGmc was *-iz not *-az.
[7] Gsg. -es after the o-stems.
[8] NApl. -i after i-stems.
The n-stems
===========
The paradigms:
masc.
Goth ON OE OS OHG
N -a -e -a -o -o
A -an -a -an -on -on,-un
G -ins -a -an -on,-en -en,-in
D -in -a -an -on,-en -en,-in
N -ans -ar -an -on -on,-un
A -ans -a -an -on -on,-un
G -ane: -a -ena -ono -o:no
D -am -om -um -on -o:m
fem.
Goth ON OE OS OHG
N -o: -a -e -a -a
A -o:n -o -an -un -u:n
G -o:ns -o -an -un -u:n
D -o:n -o -an -un -u:n
N -o:ns -or -an -un -u:n
A -o:ns -or -an -un -u:n
G -o:no: -na -ena -ono -o:no:
D -o:m -om -um -on -o:m
n. -o: -a -e -a -a
pl -o:na -o -an -on,-un -un
For PIE, I reconstruct:
PD [expected PGmc.]
N *-o:(n) *-o(:)(N)
A *-onm. *-anuN
G *-énos *-enaz (*-iniz)
DL *-éni *-ini
N *-ones *-aniz
A *-onn.s *-anuNz
*-énn.s *-inuNz
G *-éno(:)m *-eno:N
DA *-n.bhiós *-unmiz (*-umiz)
HD
N *-é:(n) *-E(:)(N)
A *-énm. *-inuN
G *-nés *-(u)niz
DL *-néi *-(u)ni
N *-énes *-iniz
A *-nn.'s *-(u)nuNz
*-énn.s *-inuNz
G *-nó(:)m *-(u)no:N
D *-n.bhiós *-unmiz (*-umiz)
Sorted by Germanic vowel variants:
/o[:]/ /e[:]/ /a/ /i/ /u/
N -o:/-oN -e:/-eN
A -anuN -inuN
G -enaz -iniz -(u)niz
D -ini -(u)ni
N -aniz -iniz
A -anuNz -inuNz -(u)nuNz
G -ono:N -eno:N
D -umiz
While I'm sympathetic to Piotr's attempt to explain the Gothic masc. N. sg.
-a by regular sound change instead of analogy, I don't think the proposed
solution is viable. The N/WGmc forms (ON/OE -a, OS/OHG -o) show a
development exactly similar to that of the fossilized remains of the o-stem
ablative (Goth. -o:, ON/OE -a, OS/OHG -o), and given the unlikelihood of
something like *-oon in the n-stem nom. sg., I see no other alternative
than to invoke the Schleifton, by way of the PIE variant forms *-o:n
(acute) ~ *-o:~ (circumflex). Gothic has -a, not -o:, so it must derive
from something else, and it's not *-o:n (that gives Gothic -o:). Analogy
after the acc. *-an(u) is always a possibility, but my money is on a form
parallel with ON -e, from PIE *-e:n or *-e~. In the 3sg of the weak
preterite, Gothic -da/-ta does derive from someting like *-dhe:t (ON
-de/-te), so I don't see any immediate phonetic obstacles.
=======================
Miguel Carrasquer Vidal
mcv@...