Ergativity in PIE

Alexander S. Nikolaev alex at AN3039.spb.edu
Sun Jun 13 00:20:23 UTC 1999


Dear Wolfgang and List!

Thank you for your comprehensive comments; and now i shall try to
present some points "in my defense":

>I cannot fully understand this point: First, we should note that case
>marking (if ever present in a given language system) is only ONE
>(possible) feature that can become relevent regarding the location of a
>language system on the AEC (accusative-ergative continuum). Hence, it is
>difficult to infer a general behavior of a language system regarding the
>AEC from just this feature.

I'm afraid i have to disagree here. As far as i know (and you
certainly know better), there're five sections of grammar regarding
which the way of encoding A, S and P can be relevant, namely:
	1) case marking -- no comments here
	2) verbal agreement -- for ergative verbal agreement
compare the following Eskimo forms:
	k'avag' - ak'u  -  n'a --- 'i'm sleeping'
      'to sleep'-Pres.Intr.-1Sg.S.
		and
	aglat-ak'a  - t  - n'a  'they're leading me'
    to lead-Pres.Tr.-3PlA-1SgP
	3) rules of syntactic transformation
The sentence "The father kicked his son and started crying"
in an accusative language can mean only 'father started
crying after he kicked his son', the missing S automatically
being identified with A of the first part of sentense; and e.g. in
Dyirbal sentence like this could only mean 'the son started crying
having been kicked by his father' [sorry for such a brutal example]
	4) word order: in Dyirbal unmarked word order is S-Vi
	and P-A-Vtr, while in english it would be A(S)-V-P
	5) the structure of compounds

But: in a given ergative language it's not at all necessary for
the rules of ergative encoding  to be realised in ALL the sections of grammar
listed above. Thus i think it's possible to claim PIE ergativity,
even if only the ergative structure of nominal system is
reconstructed; meanwhile all of the other language mechanisms
may keep on functioning on accusative basis. Cf. Chukchee, where
the ergative system of case marking is attested, but the verbal
agreement is accusative. Or the Walbiri language, where in a
transitive sentence the 1st actant has the marker of an ergative
case, but the markers of the A in verb are the same as of S.
I'm sorry for this stepping aside into the problems of typology...

<Second, if we include other aspects of
<morphological, syntactic, and pragmatic ergativity/accusativity
<(horribile dictu), the data demonstrate that a system with a
<syncretistic ergative case marker (irrelevant of how this syncretism
<pronounces itself) can be located either more on the middle of the AEC
<or on its very "edge". E.g., some East cauacsian languages have case
<syncretism but a strong ergative agreement pattern, whereas Kartvelian
<(South Caucasian) languages mostly show an "independent" case marker
<(stemming from the deictic paradigm), but have a strong accusative AGR
<pattern (iwth some excceptions).

Noted. Still, i had really no intention to say, that the existence of
a syncretistic ergative case marker in a language *presupposes*
the position of this language in the middle of AEC, not "on the
edge". Just there is a piece of evidence, which allows to
reconstruct both syncretistic erg. case marker and the tense
split for PIE independently from each other, i.e. my logics is
NOT that one implies the other. But i really expressed this in a
clumsy way.

<Note that case syncretism in EC (East Caucasian) for instance does not
<correspopnd to an TAM split.

See the last paragraph.

<Moreover, to claim that Burushaski's
<ergativity in restricted to preterit tense(s) seems not to be confirmed
<by ALL data available [ergativity in Burushaski is a very complicated
<matter because of its polypersonal agreement system..

Regretfully, i didn't succeed in obtaining a more or less
comprehensive description of Burushaski (could you perhaps give a reference?),
but still even if the agreement system is polypersonal in all
tenses, this does not necessarily mean IMHO, that one cannot say,
that "case marking in Burushaski is built on accusative basis
in present, and on ergative basis in preterit". These are
different mechanisms of the Language (i mean nominal system and
and verbal agreement), and they are entitled to be organised on
different basises. Besides, is polypersonal agreement
necessarily a feature of ergative system?

<The hypothesis that ERG is
<confined to preterit tense(s) in Eskimo is questionable, too [see the
<agreement system, some today's present tense structures clearly stem
<from old antipassives which presuppose ergativity, etc.]

I thought, that from a point of view of purely synchronic
description the case marking in Eskimo' present tense
constructions reveal no trace of ergativity. At least my
handbooks tell so. May be you could give an example?

<Perhaps you have missunderstood the actual instantiation of the
<Silverstein Hierarchy (SH). Today, SH is regarded as a a general
<behavior of lingustic paradigms with respect to the relationship of
<degree of animacy/empathy/centrality(speech act participants), degree of
<'natural' agentivity, and (in)transitivity. SH accounts for both ACC and
<ERG strategies (there are for instance cases of languages that use ERG
<strategies for the most central/animate/empathetic participants (SAP
<pronouns) whereas the rest of the paradigm goes accusative [some kind of
<'left shift' of the SH]).

Yes, it's my fault; again i wasn't clear enough. I know, that SH
is present in all the languages, and the scale of animacy
proposed by Silverstein is really universal. The point i want to
make is that
-Francisco Villar (1983) and Alan Rumsey (1987) thought
that NP-split at animates/inanimates according to this universal
is a necessary property of all ergative languages;
-the "classical" theory of PIE ergativity claimed that
historical IE nominative in -s reflected former ergative.
-but -s in nominative is attested only by animates.
-then the type of NP-split attested in historical IE dialects is
just diametrically opposite to that suggested by SH (in the
understanding of the respected authors named above)
-my point is that in languages with TAM split there is no NP-
split according to SH (which of course does not mean, that SH is
not present in a given langauge)
-or even the left-shift may occur --- and you witnessed it in
the last sentence of your last paragraph! -- could you perhaps
give some examples?

I'm happy to refer to the work by Larry Trask (1979) "On the origins
of ergativity" (IN: Ergativity, Plank ed.) on this subject. If
you are aware of a language, which combines NP-split at animates/
inanimates and TAM-split, i shall have to withdraw the
suggestion.

<I must admit that such a functional paradigm sounds VERY strange to me.
<The problem ist that you operate on a pure functional level without
<having any formal indication that such a split of the {*es/*os/*s}
<morphology was actually active.

Well, it's a plausible way to explain the coexistense of
adnominal genitive [mainly of posession] and adverbal genitive
of agent.

<Moreover, I do not understand how a PIE
<speaker would/could have discriminated both functions.

Am i wrong in assuming, that a speaker of a language with a
syncretistic ergative case marker sometimes encounters such a
problem, if in this language there is no strict word-order? PIE
syntax is still a debatable matter; but the first position of
adnominal genitive in the sentence is [as far as i know] certified.
(the structure GS of thematic stems may witness this, since we
know, that *yo tends to occupy Wackernagel's position:
*wir-os-yo) PIE really may have had a rigid word order with an
attribute (resp. adnominal genitive) always preceding the A/S...

<In fact, case
<syncretism generally means that the morpheme in question establishs a
<functional cluster that is disambiguished (among others) by the
<semantics of a given NP, its position with respect to the
<kernel/periphery of a clause, its semantic/syntactic/pragmatic role with
<respect to a verbal frame etc. But such oppositions are not established
<by TAM (except you propose that a PIE speaker "felt" that a *s- etc.
<marked NP with present tenses was more genitive-like than with preterit
<tenses. But this is more than ad hoc!).

See the following Eskimo example:
(-m being the marker of oblique case as well as of ergative case)

Amig'ym pain'aniln'um          lisuk'a Nina
'the [one] standing by the door recognized Nina'

amig'ym  pain'a
of door  place near it
amig'ym  pain'a-ni
of door  on the place near it
amig'ym  pain'a-ni-lnuk
of door  Part. 'being on the place near it'
amig'ym  pain'a-ni-lnum
		=the same in ergative case

You see, in this complicated Eskimo sentence the native speakers
still manage to distinguish betweem the two cases, the markers of
which have the same signifier.

What i suggest, is that the meaning of the morpheme *os/*es/*s
could be described by means of functional-semantical field
consisting of two semas, which are in complementary distribution,
namely 1) adnominal genitive case marker 2) ergative case marker

< Moreover, the history of a
<genitivus-ergativus (restricted per definitionem to transitive (!)
<structures)

Ergativus -- yes, but the genitive case (or, generally, oblique)
has no relation to the transitivity of the predicate. An adnominal
possessive case can be used in any NP, in sentence with any
predicate, cannot it?

<is often related to some kind of (alienable and/or
<inalienable) possessive coupling of a verbal structure with its presumed
<agent.
<Hence, it is the genitive which plays the primary functional
<role, from which a 'ergative' case is derived (grammaticalized).
<Consequently, there wouldn't be any functional difference in the two
<paradigmatic structures you propose, and, by consequence, no need to
<establish them at all.

This is a crux of the problem, whether to regard examples like
Old Indic   patyuh. krita:   as a possessive case which is
used with verboid, or as a relics of earlier ergative case.
I have yet known no decisive argument, which would show the
impossibility of the second approach.
Really, *-to- verboids are the most ergative-like structures in
IE verbal system: they are indifferent to diathesis, which is
well-known; and so the government they may have retained form
PIE times can be an important evidence for IE genitive as
reflecting etymological ergative case. We see the same in
Classical Armenian Perfect, which is based on -l- forms, not *-
to-, etc.

Best wishes,
Alex



More information about the Indo-european mailing list