Personal Pronouns / Ergativity

[Jens Elmegaard Rasmussen]

On Thu, 17 Jun 1999, Carol F. Justus wrote:

> Dear Jens,

> Thank you for your detailed comments. The issues are those on which there
> has traditionally been a lot of disagreement and probably will be, perhaps
> as much because of differing research goals as anything. For me
> crosslinguistic definitions of categories are important for the explicit
> criteria they offer for re-evaluating received wisdom. Obviously, every
> language system will have its own genius, and comparisons are at best
> hypotheses that need constant checking as new information becomes
> available.

Sure, sounds pretty, and I'm all for it as formulated; in reality however
old wisdom is often not checked, but simply rejected without valid reason,
when a new potential facet of the picture becomes available. There is a
widespread tendency among scholars to go easy on logic if there is an
"exciting" point of subgrouping or, better still, a "Stammbaum paradox" to
be gained. As if far-reaching conclusions needed _less_ argumentative
basis than unimportant ones.

> The Hittite -hi conjugation is, of course, intimately related to the issue
> of voice. Your statemtent: "The Anatolian -hi conjugation is the IE
> perfect, period", however, may overstate your case! I wonder if you meant
> to say this in the sense 'is identical to' rather than some sense of
> 'corresponds to' or 'reflects'? [...] I think you
> meant something more to do with formal comparison.

> On the formal the origin of the Hittite -hi conjugation, I feel more
> comfortable with Erich Neu's view that it reflects a categorial prototype
> of the historical Greek and Sanskrit perfect, not that category itself, and
> that Hittite, like the other older languages, also underwent changes from a
> PIE system that did not have the categorial oppositions of any of the
> attested languages.

I meant that the forms of the hi-conjugation continue those of the IE
perfect. The alternative demands miracles - in the plural. It would mean
that an IE undivided "inactive" (fair characteristic of the category with
H2 in the 1st sg.?) split into a mediopassive and a reduplicated perfect
independently in so many languages that it can later turn up in both
guises - with a consistent difference in function too - in all corners of
the IE linguistic area. If we derive the nucleus of the hi-conjugation
from the perfect, we have the same picture all over the map. And then we
have a unitary protolanguage, so that the IE language branches can really
come from a common older stock just as archaeologists and other
researchers of realia take for granted that they do. As a descendant of
the perfect, the hi-conjugation poses no greater problems than Germanic or
Latin, or even the equation of Greek and Sanskrit. The details are
problems of standard size which are quite easily overcome if the will is
there.

[...]

> 	-mi   -hi   -ha(hari)
> 	-si   -ti   -ta(ti)
> 	-zi   -i    -a(ri), -ta(ri)

> [...] The issue seems to be the
> implication of these forms for the reconstruction of PIE. Some people
> also identify the -hi forms with thematic actives. The mappings are
> not one-to-one between Hittite and Greek or Sanskrit like they mostly
> are between Greek and Sanskrit.

The mappings are fine to me: (1) *-mi, *-si, *-ti is no problem to
anybody. (2) Perfect *-H2a, *-tH2a, *-e with added -i to create a present
tense gives precisely Hitt. -hi, -ti, -i with no problems at all, the
core elements matching the endings of Skt.-Greek etc. with great accuracy.
(3) MP *-H2a (zero-grade *- at 2), *-tH2a, *-o/*-to give the endings af Skt.
as'i:ya'/a'mam.s-i, -tha:s (with some extension yet to be identified,
similarly OIr. -cuirther), s'a'ye/s'e'te with problems no geater than
those posed by the difference between Vedic and Avestan. Are these sets
presumed to be unrelated?

> Another major issue is whether the Hittite or Anatolian system was
> more like PIE or had undergone major category losses, i.e., did
> Hittite lose a PIE perfect and have only recollection of it in the
> Hittite -hi conjugation, or did Hittite never get so far as creating
> an inflectional perfect? Before the decipherment of Hittite, Meillet
> (1908: Dialects of IE) identified peculiarities of Greek, Armenian,
> and Indo-Iranian that he thought were post-PIE. Now that we have
> Hittite and Tocharian, there are more reasons to believe him.

I'd say the opposite: Meillet never had a good case, many conclusions
being based simply on absence of data; without old data he might as well
have lumped English together with Persian. The advent of Hitt. and Toch.
which fit the classical picture of IE quite well makes one believe
Brugmann's IE all the more. The perfect is not a separate category in
Tocharian, right; but it forms a participle with reduplication and a
/-w-/, that's got to be the perfect; some of the endings match the
class.IE perfect exactly; and the vocalism may be from *-o- of the perfect
or the *-e:- of the s-aorist which have thus merged (much as pf. and aor.
have in Latin). Tocharian has a funny imperfect type with long vocalism,
'carry' representing *bher-e- in the prs., but apparently *bhe:r-e- in the
ipf. The obvious explanation is by analofy with 'was' as it used to be,
i.e. prs. *H1es- : ipf. *e-H1es-, i.e. with the augment which gave a long
/e:/ in the ipf. if 'be'. Hitt. arhi, erweni 'come' becomes regular if the
reduplication of the perfect is remembered: *H1e-H1or-/*H1e-H1r- yields
*o:r-/*e:r- and sets a perfect model for other verbs.

> On Hittite kuen-, I translate it 'strike', not 'kill'. In Hittite royal
> annals, a king often 'strikes' the enemy with the result that sometimes the
> enemy dies, sometimes he just runs away. The action type of kuen- is not
> clearly transitive in any telic sense. Hittite kuen- is also a -mi verb for
> which the passive is the suppletive -hi verb ak(k)-/ek(k)- 'die', which is
> attested with medio-passive endings. One might have expected kuen-, if it
> behaved like English 'kill' to have active -mi and medio-passive forms.
[...]

Not if it was replaced by ak(k)- 'die' in that meaning (thus Puhvel in the
introduction to the entry ak[k]-), which incidentally only makes sense it
kuen- is 'kill', at least some of the time.

> Yes, Ilya Yakubovich thinks that there is a Sanskrit stative that argues
> for a PIE stative. In the same volume I gave reasons for a very different
> view. I stand by my reasons and would agree with Jamison's having given up
> the idea.

For the record: I see on re-reading that Jamison had it from Insler who is
then alone to be blamed or credited. I cannot accept that e:-statives are
post-PIE. Their allomorphy obeys subtle rules of IE ablaut which a late
creation would not. And, of course, they turn up all over the place, in
fact I don't know of a branch that does not supply examples. Your
"reasons" are a guess at a functional split of 'hold' to 'take' and
'have', leaving the detailed repetition of the alleged process in
different branches to chance or areal influence - and apparently
disregarding the colossal evidence for *-eH1- as a PIE derivative verbal
type. Your siding with Jasanoff ("no") against Watkins ("yes") in the
question of antiquity of *-eH1-verbs is vitiated by the illogical
conclusion drawn by Jasanoff from his - in itself sound - dismissal of
Cowgill's analysis of Goth. -ai-/-a- (habaith, haband) as from
"*- at 1-ye-/*- at 1-yo-", the dismissal being based on the correct observation
that IE appears not to have syllabic shwa before (or indeed after) /y/;
the illogical conclusion is that this does not exhaust the possibilities,
not even the obvious ones: a levelled *-eH1-ye/o- works just fine, and
even makes the Gothic stem formation identical with that of Latin (which
latter point however may be coincidental). - Yakubovich' reasons for
accepting a stative behind some Skt. verbs in -a:-ya- (oddly segmented
"-a:i-a-" in the title) are possible cognates, some looking quite good to
me. But even without that, there certainly is a stative in Ved.
sana:ya'nt- 'being old' which must be *sen-e/o- 'old' + (zero-grade of)
*-(e)H1- 'be' + prs.-forming *-ye/o-, just as Lat. sene:sco: is *sen-e/o-
+ *-(e)H1- + inchoative ("s-aorist) *-s- + prs.-forming *-ye/o- (with the
s-aorist morpheme replaced by, or developing into, *-sk^e/o-), which
latter is the durative ("prs.-stem") variant of Hitt. -es- of ingressive
verbs ('become -'). And that makes the stative, no matter what its
ultimate origin, a PIE derivative type.

> There are many issues here, the last more interesting to me:

[snip of quote and part of answer, for it's summed up by:]

> The real challenge is to try to arrive at criteria for identifying the
> 'nucleus for the expansion' as opposed to the layers that got added.

> Contributions to this issue, however, are more likely to come in articles
> than our current format. []

That's your choice which I respect. However, the list could make progress
in a matter of days, while articles take decades to work.

Kind regards,

Jens