Personal Pronouns / Ergativity

Tue Jun 1 10:54:40 UTC 1999

Peter wrote:

> A couple of features explained by a theory of ergativity are:
> (a) the identity of nominative and accusative for all neuters.
> (b) the lack of a true original passive.
> (c) the origin of the verbal endings (see Szemerenyi p 330 for a brief and
> confusing summary)

[Let me first ask the audience to excuse the format of my lasting
posting [Re: accusative and ergative languages (May 28)]: I accidently
hit the "send" bottom before having finished my mail and before having
proof read the text [which is - viewing my bad English - a necessary
step - sorry if some typos remain in THIS mail, too!]].

Now, let's turn to Peter's criteria for possible IE features of
ergativity and check them with respect to what General Linguistics and
Typology tell us about ergativity etc. AS I said in my last posting, a
general definition of ACC vs. ERG should be based on the assumption that
these phenomena do not represent categorial properties of a given
language system, but reflect the behavior of certain (compatible)
morphosyntactic and morphosemantic structures with respect to a) system
internal relations, b) communicative conditions, and c) cognitive aspect
of information processing and categorization. These parameters interact
in a complex way and cannot be seen in isolation (except for analytic
purposes); moreover, all of them are embedded into the historical
dimension of language systems which renders a purely synchronic
interpretation of such morphosyntatic features highly problematic.

The general behavior of ACC and ERG can be formalizes as {S=A;O} for
ACC and {S=O;A} for ERG which read: In an ACC stategy, the structures
associated with the "grammatical" roles subjecective (S) and agentive
(A) behave alike, wheras the objective (O) (clustered with the indirect
objective (IO)) behave different. In an ERG strategy, it is S and O that
behave alike, wheras A behaves different. I think, this formulation is
on common ground. We should be aware of the fact that the above given
formulae do not refer to a specific linguistic category such as NP or
AGR. In my mind ACC and ERG strategies - often secondarily splitted up -
dominate most of the the 'operating system' of a language that is those
parts of grammar that control the linguistic interpretation of event
experience in terms of 'sentences'. However, some of these linguistic
categories are more likely to be effected by the accusative-ergatiev
continumm (AEC) than others which can be explained by the assumption
that the totality of grammar is prototypically organized: The more
central a part of grammar is the more likely it is that it plays a
crucial role with respect to the AEC.

A problem of locating reconstructed paradigms on the AEC surely is that
we do not deal with a homogenous 'synchronic' stratum but with different
time layers: Parts of our reconstructed IE grammar may be relatively
young, others much older (depending on the comparative evidence and on
the time depth reached by documented sources).

Let us check now some of the reconstructed paradigms of IE (we deal
with some kind of internal functional reconstruction here):

> (a) the identity of nominative and accusative for all neuters.

It should easily come clear that this feature in itself does not have
anything to do with the AEC. The only thing we can conclude from this is
that true neuters rarely show up as agents (that they hardly ever played
the role of a subjective (S) or agentive (A) except in a metaphorical
sense). Neutral NPs hence are very likely to represents objectives (O).
This is very natural - it tells us nothing about the AEC just because
this feature does NOT refer to the discrimination of S, A, and O. What
we get is:

(1)
	NP[+neutral < -animate]
	>
	{O}

But how NPs reflected S and A? Accrding to MY klnowledge there is NO
evidence that intransitive NPs (i.e., S) ever behaved different from
transitive NPs (i.e, A) with respect to CASE marking. Note that this
does not touch the question of case marking itself! For nominal (!) NPs
we get:

(2)
	NP[+animate]
	>
	{S=A}

[Note that I use 'animate' in a rather vague sense - I neglect a
discussion of which underlying semantics of this cut-off point we should
ascribe to IE]

The problem of IE case inflection surely is that is seems rather
atypical in a typologcal perspective (though NOT unattested elsewhere!).
One point is that S and A both are marked or zero (as it is (lateron)
the case for O, cf. (I neglect the dual here):

(3)
	S
	[SG: O, *-s; PL: *-es]

	A
	[SG: 0, -*s; PL: *-es]

	0
	[SG: 0, *-m; PL: *-H2]

The O-marker *-m seems to be a somewhat younger formation stemming from
a directive or so that secondarily effected the neutral o-stems. The
rise of O-marking via a directive (or, let us say, any 'locative') case
is a typical strategy that has to be associated with the ACC strategy.
Normally, it is labeled 'Differentiated Object(ive) Marking (DOM)' or
simply O-split. It can be paralleled with the O-plit for instance in
Spanish. In late IE, the marked variant of the O-split obviously had
already been generalized (IF PL ACC *-ns stems from *-ms and this again
from *-m-s, then the plural *-s-marker should be regarded as a
relatively young phenomenon). For an earlier phase of IE NP inflection
we can assume (I neglect the plural forms]:

(4)
	S
	[O, *-s]
	INFER
	[+animate]

	A
	[0, *-s]
	INFER
	[+naimate]

	O
	[0]

	INFER
	[±animate]

Evidently, this paradigm does not reflect any kind of ERG strategy. Such
a strategy often has been inferred from the fact that the {S=A}-marker
*-s seems to have something in common with the genitive (singular (!))
(*-es, *-os, *-s). From this some kind of 'genitivus-ergativus' had been
reconstructed from Pre-IE. Naturally, a genitivus-ergativus is attested
in a considerable number of languages (Yupik-Eskimo, Lak (East
Caucasian), to name only two). However, again IE *-s does not behave
ERG, even IF we can associate it with the genitive marker (singular):
*-(e/o)s also encodes the S-function, which is ANTI ERG. I think it
would be much better to propose a strategy of topicalization within the
ACC (or 'neutral'?) paradigm for Pre-IE, cf.:

(5)     *NP[S]-s VERB < *NP[S]-s[TOP] VERB

	*NP[A]-s NP[O]-zero VERB < *NP[A]-s[TOP] NP[O]-zero VERB

It would then be very temptive to relative *-s to the most natural means
to encode topicality, namely to the *so- deixis (in {S=A}-function).
Hence in Pre-IE a phrase like 'the woman went' would have been
'WOMAN(*-s) went', and 'the woman saw the tree' would read 'WOMAN(*-s)
tree(-zero) saw' (lateron the transitive structure would have been
changed to 'WOMAN(*-s) to-tree(*-m) saw').

Those NP that were zero-marked in {S=A} function obviously carried some
kind of inherent topicality which did not necessitate an *-s-marking. To
conclude this point: The noun inflection of IE does not show ANY trace
of an ERG strategy, rather we have to deal with an ACC based
topicalization that has its good parallel for instance in some
Afro-Asiatic languages...

> (b) the lack of a true original passive.

This again has nothing to do with the AEC! If IE lacked a true passive
the only thing we can infer from that is that IE once had been a role
dominated language (see VanValin's 'Role-and Reference Grammar') that
did not use fore- or backgrounding strategies. There are ACC languages
with and without passives just as ERG languages with and without
antipassives and/or passives...

> (c) the origin of the verbal endings

We have to distinguish 'origin' from 'function'! In a functional
perspective, IE personal clitics ALWAYS show an ACC behavior: They are -
as far as I know - NEVER conditioned by the O-role ('object', if you
want). In order to clarufy this point cf. the following example from Lak
(East Caucasian):

(6)
	ta-na-l ZERO-at:-ay-s:a-ra na

	he-SA-ERG CLI-hit-PRES-ASS-1Sg I:ABS

	'He surely hits me'

[SA = stem augment, CLI = noun class I [+masc;+hum], ASS = assertive]

Here, ZERO as well as -ra are triggered by the first person pronoun 'I'
in objective function [note that the system of personal aggreement in
LAk is much more complicated, so please do not infer from this example
that Lak ALWAYS has O-AGR!). Nothing the like is known from IE. If we
again look at my above given TOP-hypothesis, it comes clear, why: The
clitic in the verb obviously marks the anaphoric slot produced by the
TOP procedure, and this is ACC, cf.:

(7)
	NP(S)(*-s)[TOP] VERB-AGR[TOP]

	NP[A](*-s)[TOP] NP[O](-zero) VERB-AGR[TOP]

Note that (7) does not tell us about the formal history of the
AGR-elements.If we look at the typology of the grammaticalization of
AGR-paradigms, it becomes obvious that such paradigms hardly evolve at
once: Rather, they start with one person (very often 1Sg) or the
SAP-cluster (1Sg and 2Sg encoded by ONE morpheme) and gradually become
generalized. For Pre-IE it seems probable to regard the 1Sg as a
starting point: But IF the 1Sg marker is the grammaticalized form of the
1Sg personal pronoun (*-m etc.) then this does NOT force us to look for
the same source regarding the other persons (esp. 2Sg.). It may well be
that e.g. the 2Sg *-s stems from say a deictic paradigm...
But again this is irrelevant for the AEC as long as AGR behaves in one
direction, namely ACC in IE (the perfect(ive) markers do not change the
picture).

We could go one discussing the paradigm of personal and deicitic
pronouns, traces of the discourse cohesion devices in IE, relativization
strategies, word order etc. (what I won't do here). Whatever that basic
strategies of IE morphosyntax and morphosematics had been: Neither its
'operating system' in its globality nor the relevant subparadigms show
any convincing trace of {S=O;A} behavior [in case you know of one plase
tell me!]. This does not mean that I regard ACC as being more 'natural'
than ERG or so: both are two parallel options to which language systems
refer to a different extent. What is surprising with respect to IE is
the fact that it seems to have been quite radically ACC dominated. But
explaining this finding is another story...

Wolfgang
--
[Please note new phone number (office) :+89-2180 5343]
___________________________________
| Prof. Dr. Wolfgang Schulze
| Institut fuer Allgemeine und Indogermanische Sprachwissenschaft
| Ludwig-Maximilians-Universitaet Muenchen
| Geschwister-Scholl-Platz 1
| D-80539 Muenchen
| Tel:	+89-21802486 (secr.)
|	+89-21805343 (office) NEW ! NEW !
| Fax:	+89-21805345
| Email: W.Schulze at mail.lrz-muenchen.de
|
http://www.lrz-muenchen.de/~wschulze/
_____________________________________________________