Personal Pronouns / Ergativity

[Wolfgang Schulze]

[ moderator re-formatted ]

In response to my last posting Eduard Selleslagh said:

> As you well know German uses such a genitivus-ergativus case with the
> passive (and in other constructions, to point at the author), albeit
> periphrastically, with the preposition 'von', which could just as well be
> called 'ablativus-ergativus', the genitive and the ablative both being
> 'pointers' to the origin of the action. Note the similarity of the Latin
> construction : passive + a(b) + ablative. So, it seems that the need for
> some (pseudo-?)ergative way of speaking is still lingering on in IE
> languages.

As Leo said: "[]hat's pushing it a bit far." The constructions Ed
mentioned belong to the paradigm of fore/backgrounding strategies (aka
"diathesis"). Diathesis presupposes that there is a basic strategy of
enconding participants in a "simple sentence", passives and/or
antipassives. The way how backgrounded participants (most often semantic
agents) are marked heavily depends on the individual language though
some general tendencies may be observed (among them a
genitivus-ablativus etc.). The most characteristic feature of most types
of diathesis is that such marked NPs are normally located in the
periphery of a "sentence", indicated by an appropriate oblique
construction. Note that such peripheries sometimes can be secondarily
splitted up, e.g. in German (cf. Leo: "German also uses _durch_
'through' for this
purpose, mainly for less active participants." [I do not in total agree
with the semantics proposed by Leo's, but that does not matter here].
The question of whether or not a language system operates via diathesis
depends on the general nature of role assigment in the language: We have
to assume that NPs are marked by a functional cluster that covers (at
least) the three domains 'semantics', 'reference', and 'pragmatics'
[semantic, syntactic, and pragmatical roles, if you prefer more
traditional terms]. Some languages favor a semantic-centered clustering
(such as our famous Dakota), others reference-centering (many languages
with classical diathesis), again others pragmatic centering (Tagalog
etc.). Diathesis as a referential strategy (phrase internal as well as
discourse dependent) presupposes a reference-dominated type. Many Modern
IE languages are of this type (though some of them show a shift twoards
pragmatic centering (among them German and Dutch), but PIE itself
obviously lacked this strategy which comes clear from the fact that we
cannot reconstruct a common "passive" for PIE. Diathesis is are very
active feature in language change. It can come and go, and nothing
allows us to propose a Passive for PIE just because a number of (modern)
IE languages share this feature.

If we look at the question of ERG and passive, we can sometimes observe
a tendency to establish a pseudo-ergative strategy based on the passive
(some modern Indo-Iranian languages, partly Cl. Armenian etc.). But this
a secondary process often bound to specific (perfective) TAM forms. And:
The ergativisation of passives (including shifts in word order etc.)
does not argue for an earlier phase of ERG, on the contrary: Passives
are typical for ACC systems (though also attested in what sometimes are
(wrongly) thought to be "true" or "pure" ERG systems such as Inuit). The
standard diathesis in ERG systems is the antipassive, cf.

	ACC	PASS	ERG	AP
	S=A	[A]	S=O	[O}
        O	S(<O)	A	S(<A)

Now, IF (i say IF) you want to ascribe some ERG features to the PIE
system it would make much more sense to declare ALL transitives as old
(and generalized) antipassives, cf. (I use -s and -m  as symbols for
case for here):

		A-s O-m VERB-AGR(A)
	<	S-s [O-m] VERB[AP]-AGR(S)
	ERG	A-ERG 0-ABS VERB-AGR[O]

But this does not make sense as long as we don't have substantial
evidence for ERG strategies elsewhere in the paradigm (e.g. in Georgian
which has an ACC (< AP) strategy in the present/future tenses/modes, but
an ERG strategy still persent in the aorist, hence Georgian mirrors say
some Iranian languages on an ergative basis). In PIE, there is no such
evindece, I think (also because PIE seems to have been role- and NOT
reference dominated: Consequently, PIE probably could not have AP
strategies, just as it lacked PASS strategies.

Peter commented upon another aspect: I argued 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)". He wrote:

> Perhaps I fail to understand.   Did you mean to count both subjective and
> agentive as "agents", and say that it was natural that inanimates should not
> play that role?   If pre-IE had an animate / inanimate distinction, we can
> believe that inanimates might never or very rarely be agents, but it is more
> difficult to believe they were never subjects.   I find it difficult to
> believe that in a language which produced PIE, Mr IE was unable to say to
> Mrs IE, "The <inanimate thing> is nice (or dangerous, or whatever)."

This surely is a problematic aspect. What we need is a better
characterization of what is "subject", "subjective", "agentive" and so
on. I will skip this question here because it would lead us into the
deep ocean of Grammar Theory which has as many faces as researchers, I
think. In short: True statives very often show an atypical behavior with
respect to the AEC which in itself is a question of how the degree of
transitivity of
DYNAMIC actions is formalized [ERG focusses on the "effect" of an
action, ACC on the "source" etc., see among MANY others Hopper/Thompson
1980, Silverstein 1976 (and Schulze 1998, if I may humbly add this
reference). Sometimes, statives appear to be derived from intransitives:
They are either subsumed under this strategy (as in many ACC and ERG
systems) or split of there from in terms of Split-S marking ("active
typology"). The fact that statives and perfective structures seem tio
have been closely related in PIE indicates that PIE knew (at least in
its agreement system) some kind of Split-S. IF this split was present
with case marking, too, then we should assume that (in)animate statives
were zero-marked, just as the old O-marker before the O-Split via *-m
became effective, cf.

S(inactive)-zero VERB(stative)(AGR(S(inactive)))
S(active)-s VERB(AGR(S(active)))
A-s O-zero/m VERB(AGR(A))

Note that I do not ascribe a general active typology to PIE (as Lehmman
and others), I which I do not believe. It is just a simple S-split which
is as "normal" as O-splits are. As our moderator has put it: "The
reference is, I believe, to o-stem neuter nominative/accusative which
looks like o-stem non-neuter accusative.  However, in other stem
formations, the parallel in appearance does not exist". Old neuters were
zero-marked both in S and O function, whereas animates were -s-marked in
S- and -m-marked in O-function, cf.

		Animate		Inanimate
S(inact) 	-s		-0
S(act)		-s		---
A		-s		---
0		-m		-0

[A "good" acticve typology would yield:

	 	Animate		Inanimate
S(inact) 	---		-m
S(act)		-s		---
A		-s		---
0		-m		-m  (if we use the PPIE coding system
                                     as a symbol)]

Hence, I think that PIE showed a doubled split in its case marking:
S-split and O-split, before both splits were harmonized again. To make
this point clear, I'd like to add what Mr. or Mrs. PIE would may have
used as a scheme for case marking and agreement (I trasnlate *-s into
TOP[ic], *-m in DIR[ective], and *-0 (zero) into ZERO:

a) water-ZERO cold-AGR(water(inact))    S(inanim) = ZERO
b) woman-TOP young(-AGR(woman(active))  S(anim)   = TOP
c) woman-TOP run-(AGR(woman(active))	S(anim)   = TOP
d) woman-TOP man-DIR see-AGR(woman)	A(anim)   = TOP
					O(anim)   = DIR	
e) woman-TOP water-ZERO see-AGR(woman)  A(anim)   = TOP
					O(inanim) = ZERO

Naturally, such as scheme is nothing but a very rough and only tentative
paradigm. We have to assume that PIE knew as many (esp. metaphorical)
variants as documented on other languages. Such variants may have
allowed PIE spekaers to use inanimates as S in dynamic constructions or
as A. Still, the overall picture remains the same: The operating system
of PIE clearly showed an ACC strategy in its protypical kernel,
semantically splitted according to [±animate] or so. This ACC strategy
seemed to be dominated by topicalization routines with animates, a clear
indices for the semantic basis of PIE "case" marking. Finally, AGR does
not change this picture, even if we assign the *-H2e etc. series to
statives/inactives, and the *-m etc. series to dynamics/actives: In this
case, even the dichotomy [±anim] becomes irrelevant, because it does not
show up in a specific set of clitics. ALL these clitics have an ACC AGR
scheme...

I hope, that this helps to understand my point.

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/
_____________________________________________________