On Fri, 27 Jun 2003 21:52:15 +0100, Richard Wordingham
<richard@...> wrote:

>How does the 'r' of the Armenian cognate (at least of the IE forms) asr fit
>into this? Does it at least relate to the -r- of Latin pecus despite what I
>thought?

Armenian has a whole bunch of those unexpected -r's:

Adjectives like barjr /bardz&r/ "high", canr /tsan&r/ "heavy", manr /man&r/
"small", p`ok`r /pHokH&r/ "little", k`aLc`r /kHal~tsH&r/ "sweet", goLtr
/gol~t&r/ "tender", t`anjr /tHandz&r/ "thick".
Nouns like asr /as&r/ "fleece, wool", caLr /tsal~&r/ "laughter", meLr
/mel~&r/ "honey", artawsr /artO:s&r/ "tear", cunr /tsun&r/ "knee".

As far as the eymology of these items is clear, they correspond to IE
u-stems (*bhérg^hus, bhr.g^hú- > barjr "high", *men(h)us, mn.(h)ú- > manr
"small", *swah2dus ~> k`aLc`r "sweet", *peku, pkú- > asr "wool", *medhu >
meLr "honey", *drak^u- > artwasr "tear", *g^onu- > cunr "knee"), except
caLr "laughter (Grk. gélo:s), an s-stem ending in -o:s (Arm. > u). The
nouns are moreover _neuter_ u-stems.

The case forms are also very interesting. In the singular they are
ordinary u-stems (except for the -r in the Nom/Acc):

NA barjr asr
GDLI barju asu

In the plural, however, the nouns are a:-stems (no doubt because of the old
n.pl. *-u-h2 > *-wa):

N artasuk` < *drak^wa + -k` cungk` < *g^onwa + -k`
AL artasus < *drak^wa + -s cungs < *g^onwa + -s
GDAb artasuac` < *drak^wa + -isk^- cungac` < *g^onwa + -(i)sk^-

The adjectives are n-stems in the plural:

N canunk` < *g^n.nones barjunk` < *bh(e)rg^hones
A canuns < *g^n.nonn.s barjuns < *bh(e)rg^honn.s
GDAb canunc` < *g^n.non- + -isk^- barjanc` < *bhr.g^hn.- + *-isk^

Attempts to explain the -r in the NA sg. by an appeal to the RUKI-rule
(*-us > *-us^ > *-uz^ > *-ur > -r) fail in the face of the fact that the
nouns are neuter, so never had an -s in the nom. sg.

The fact that the nouns are neuter suggests another well-known
irregularity: that of the r/n-heteroclitics (*-n > *-r, *-n- stays).

My solution is that these words were originally *un-stems. For neuter
nouns, this would have given:

NA **gá:n-un "knee"
G **ga:n-ún-a:s

Applying the usual soundlaws (*á: > *o; pretonic *a: > é, posttonic *a: >
o, *-n > *-r, *ún > *ew, zero-grade), the expected result would be:

NA *g^ónur
G *g^énwos

(The variant **gá:n-unt, **ga:n-ún-tas would have given *g^ónur, *génwn.ts
= Greek gónu, goúnatos < *gonwn.tos < *genwn.tos).

It is not surprising then, that the Armenian adjectives are n-stems in the
plural. The paradigm would originally have been:

N *bhárghun-z > *bhérg^ho:n (-u:n?)
A *bhárghun-m > *bhérg^hum
n *bhárghun > *bhérg^hur = barjr (for *berjr)
G *bharghún-a:s > *bhr.g^hwós = barju
pl.
N *bhárghun-es > *bhérg^hunes = barjunk` (for *berjunk`)
n. *bhárghun-h2 > *bhérg^o:r (-u:r?)
D. *bharghun-bhiá:s > *bhr.g^hn.bhiós = barjan-c`

In Sanskrit, the adjective madhu- "sweet", for instance, makes neuter I.
mádhuna:, D. mádhune:, G. mádhunas, L. mádhuni, NA pl. mádhu:ni.

In Tocharian, u-stem adjectives have been replaced by ro-stems (pärkäre
"high" = Arm. barjr, swa:re "sweet" = Arm. k`aLc`r).

In Latin, pecus, pecoris, and pecunia, also show traces of the old *n (~
*r).

Another old *un-stem is probably *dá:r-un > *dóru(r) "tree, oak" (Skt. gen.
drún.as).

=======================
Miguel Carrasquer Vidal
mcv@...