German ge- ptcpl cognates?

[Sean Crist]

On Tue, 25 Jan 2000 ECOLING at aol.com wrote:

> I was just recently contemplating the augment *e- which precedes
> certain completed-action verbs in Indic and in Greek, and which, I
> gather from one correspondent, is usually taken as an inheritance from
> some common stage, one of several manifestations of a close
> Greek-Vedic relationship.

> If one takes that point of view, then it implies a Greek-Indic common
> innovation compared with PIE. Is that branch on a tree supportable?
> (differs from UPenn, right?)

> If not, then must one take it as an inheritance from PIE, lost elsewhere?

Represented in a flattened form, the tree computed by Ringe, Warnow and
Taylor is as follows:

(Anat (Toch ((Ital Celt) {(Gk Arm) (BS IIr)})))

(Notice that I've left Gmc out in this version; in earlier versions it was
grouped with BS and IIr, but the placement of Gmc has turned out be a
bigger problem tham expected.)

The augment is found in Greek, Armenian, and Indo-Iranian (and Phrygian,
which wasn't included in the tree).  There is no evidence for the augment
prefix outside of the grouping which I have enclosed in the curled
brackets.  Thus, the augment could have been innovation in
proto-Gk-Arm-BS-IIr, and then have been lost in Balto-Slavic.  A
morphological loss of this kind is not surprising.

> it occurs to me to wonder about German ge- of past participles, which
> (with loss of /g-/) shows up also in the "e" of the English form
> "enough", related to German "genug", from o-grade of a verb *nek- 'to
> reach, attain'.  What is the origin of that prefix in German?

> Is it just barely conceivable that it might be related to the Sanskrit
> and Greek augment, and that it began with a laryngeal? (I am fully
> aware that affixes do not always follow the same sound changes as do
> roots - but also very hesitant to posit an otherwise unproven
> irregularity of sound change, so would want some pretty good
> demonstrations of cognate functions, at least ones which could develop
> from a common origin. I think the functions of German <ge-> and of the
> Greek and Sanskrit augment, in completive contexts, are highly
> similar.

> Could the origin have been something like this? (or with a different
> vowel, reduced to /e/ as for many other German unstressed verbal
> prefixes, then generalized ...?)

> *He-

> ??

> Pardon, I am not a Germanicist and have no immediate access to
> something that would tell me. Pokorny's Comparative Germanic Grammar
> pp.205-206 states a relation to Latin co(m)-, but such a hypothesis is
> to me much more improbable on semantic grounds.  In this view, I would
> assume, the /gV-/ prefixes gradually spread from their point of origin
> at the expense of other prefixes.  Perfectly possible. Was that
> hypothesis posited long ago for simply for lack of anything better, or
> because <ge-> shows up only in some of the western Germanic languages
> (OHG, English, etc.)? Or is there substantial support behind it, such
> as details of the gradual stages of infiltration from Latin?

Your speculation that there might be a connection between the augment and
Germanic ge- is a natural thing to wonder.  I know because I wondered this
myself, but when I asked Don Ringe about it, he said that there is no
connection.  Actually, you are in luck, because I can quote the email he
sent to me on 27 May 1997 (I'm sure he wouldn't mind me quoting this
private email):

(I asked something like, "I guess there no connection between Gmc. ge- and
the augment?")

> Sean-- Right, none whatever.  The "augment" (= marker of past tenses)
> has clear cognates in Indo-Iranian, Armenian, and Phrygian, but not
> elsewhere. The fact that there are no clear cognates in Anatolian or
> Tocharian (or Italic or Celtic, for that matter) could mean that it
> isn't PIE and was just an innovation of the central group; that would
> still mean that pre-Germanic once had it, though, and lost it.

> German ge- is < PGmc. *ga- (cf. Gothic ga-); actually appears to be
> cognate with Latin com-/con-/co-, to judge from equations like Goth.
> gamains = Lat. commu:nis (and it *was* an i-stem in PGmc.:  note
> *gamainiz --> pre- OE *gamainijaz > *gaema:ni: (where "ae" is the low
> front vowel) > OE gemae:ne, with umlaut) and loan-translations like late
> Lat. compa:nio: --> Goth. gahlaiba (pa:nis = hlaifs 'bread').  The
> phonology is screwed up, though; in particular, it looks like Verner's
> Law has applied to a word- initial unstressed syllable, which is frankly
> weird.

To answer a few of the other points you bring up:

-You suggest that Gmc /g/ could be the reflex of a laryngeal in ge-.  In
cases where we're lucky enough to have evidence from the other branches
for a word-initial laryngeal, Germanic uniformly has zero (minus a very
technical point regarding the development of the Gmc strong verb classes
which isn't relevant here).  If a laryngeal developed into Gmc */g/ in
this case, it would be the only such case we have, and it would be
inconsistent with all of the other Gmc words descending from a PIE word
with an initial laryngeal.

-Also, if there were a laryngeal, it would make its presence felt in Greek
and Indo-Iranian.  It's too late in the evening for me to go digging thru
my notes, but a laryngeal would affect the vowel quality in Greek, and if
I remember right, a larygeal usually comes out at *a in Indo-Iranian.
I'm not sure what happens to a word-initial larygeal before a vowel in IIr
(I do seem to remember that word-initial augment + HC- > a:C, i.e. where
the laryngeal belongs to the stem; but I'm not sure if HaC- would give aC-
or a:C-).

-As Don alludes to, the meaning of *ga- in Germanic has changed over time
(consider the German nouns _Gebirge_, _Geschirr_, etc., where the meaning
of ge- is "collective", not "perfective"). .  It appears that an early
meaning was very much like that of Latin co(m)-, roughly "together".  If
you want to claim that *ga- and co(m)- are cognate, the semantics at least
will work out.  However, as Don points out, the ordinary phonology doesn't
work out for co(m)- to be cognate with *ga- (PIE short *o _would_ give
PGmc *a, so the vowel checks out; but the consonant doesn't work.)

I'm a little surprised that Don accepts the suggestion that co(m)- might
be cognate with *ga-; the claim looks weak to me. It's true that *ga- is
often used in forming calques from Latin words with co(m)-; but this
doesn't strike me as a particularly good argument that the two are
cognate.  You can certainly use a word or affix in forming a calque
without the word or affix being cognate with anything in the language that
you're calquing on.

Personally, I think we need to say that we just don't know where Gmc. ge-
comes from.  In any case, it definitely can't be cognate with the augment,
because there would be much worse phonological problems with that claim
than with the claim that Gmc. *ga- is cognate with com-.

  \/ __ __    _\_     --Sean Crist  (kurisuto at unagi.cis.upenn.edu)
 ---  |  |    \ /     http://www.ling.upenn.edu/~kurisuto/
  _| ,| ,|   -----
  _| ,| ,|    [_]
   |  |  |    [_]