accusative and ergative languages

Wolfgang Schulze W.Schulze at
Tue Aug 3 15:36:42 UTC 1999

Dear Pat and IEists,

> Pat writes:

> Since "discourse cohesion strategy" is not defined in Larry's dictionary, I
> have no idea exactly how one will want to define it. Perhaps 'anaphors' are
> excluded; perhaps not.

This really is funny! Do you mean that linguistic categories,
paradigmatic or syntactic strategies, or other structural items terms of
which are not defined in Larry's dictionary do not exist? Stefan's
'discourse cohesion strategies' (better known as 'discourse cohesion
devices' (DCD)) represent a well-established label in typology (as well
as in CA, by the way). DCDs are those linguistic structures
(suprasegmental startegies, morphemes, syntatic conventions) or emergent
'activities' of related structures that allow the coupling of phrasal
information and its co/context (you probably know the difference - you
can also say 'endophoric' = cotext, 'exophoric' = context). The main
point is that besides the grammaticalization of elements in terms of a
DCD (for instance, topic marker, co- vs. switch reference markers, etc.)
sentence internal strategies may be used to generate discourse cohesion.
The most prominent one is the notion of 'pivot' which in some respects
shares properties with 'subjectivization'. How sensitive actance marking
is with respect to discourse cohesion can be easily shown, let me quote
an well-known example from Hua (Papua):

(1)	d-go' ro-0-da Kori-ue
	ISg:O-scare feel-3Sg:A-1Sg:SS run=away-1Sg:S
	'I felt scared and ran away.'

(2)	Korihu-ga-na d-auia-h-ie
	run=away-1Sg:DS-3Sg 1Sg:O-shame-feel-3Sg:A    (Haiman 1980:360)
	'I ran away, and I am ashamed.'

[SS = same 'subject', DS = different 'subject', A = agentive, O =
objective, S = subjective]

There is a kind of 'impersonal' cionstruction in Hua (often related to
verba sentiendi or uncontrolled verbs) that has an 'un-accusative'
reading (objective + verb). In (1) this structures precedes the
intransitive sentence 'I ran away'. The expected switch from SS
(conditioned by S=A) to DS (here O followed by S) does not take place.
In (2), however, just this switch happens. Here, S precedes O, and we
have DS! SS/DS serve as a DCD here, their sentence internal
distribution, however, is conditioned by aspects of the information flow
within the co(!)text.

The same is often true for accusative or ergative strategies (cf. German
'ich schlug den Hund und wurde von ihm gebissen' ('subject' as a DCD)
vs. 'ich schlug den Hund [und] der [Hund] biß mich' (switch reference,
which generates a totally different pragmatic or discoursive strategy...

Another nice example in from Khalka-Mongolian (Stefan could tell you
much more about that, he's the expert in Mongolian, not me! and sorry
for the very rough transcription/transliteration), cf.

(3)	cecg-iyg gertee xare/zh yav-ax-ad ne/ genet Baatar dayralda-zhee
	Zezeg-ACC at=home back go-VN-LOC 3Sg:TOP suddenly Baatar
	'As Ceceg was going home, Baatar suddenly meet him.'

(4)	Bi dund surguule/ tögs-ööd end ir-sen
	1Sg:NOM middle schol:ACC:indef finish-CV:PERF PROX come-VN:PRET
	'After having fished the middle school, I came here.'

Here, DS conditions the switch from S=A encoded by the nominative (as in
(4)) to S=A encoded by the (definite) accusative (cecg-iyg) in (4). You
see that sentence internal coding strategies (incl. ACC/PASS and ERG/AP
strategies) may be very well dominated by the 'discourse', or, in other
words, funtion as a DCD. We cannot explain ACC and ERG behavior, if we
do not refer to such possible strategies which sometimes are even
primary as opposed to secondary sentence internal reasons emerging from
discourse cohesion.... We must not rule them out just because they are
not mentioned in a teminological dictionary...

In another posting, Pat says:

> I suspect the (original [?]) real function of an
> anti-passive is suggesting lessened effective agency. What do you think?

It's the same story as with passives: foregrounding or backgrounding
(often intertwined)? As I said: Antipassives reflect nothing but a
morphosyntactic (and somtimes discoursive) behavior. They effect the
relationsship of A and O just as pssaives do. The formal 'demotion' of
A=ERG to A(>S)=ABS and that of O=ABS to O(>PERIPH)=OBL (OBL serves as a
dummy here) may be caused by the semantics of A as well as that of O. If
we have an A-motivited antipassive than aspects like 'lessened
agentivity' MAY enter the game (remeber that this presupposes that A has
a semantic reading here. We all know that A is structural term, which
MAY be spelled out semantically, but which also may serve as a
syntactic, and/or pragmatic strategy of grounding and/or discourse
cohesion, see above). If we have an O-dominated antipassive, then
parameters such as mass nouns and non-countability, lessening of
referentiality (for instance in Yupik) etc. may become relevant (the
general tendency is that A is more likely to react on semantic settings,
and O to react on pragmatic settings (both can be metaphorized SEM>PRA
for A and PRA>SEM for O). Finally, the verbal relation itself may
condition antipassives, for instance durative, habitual, imperfective,
etc. All depends on the individual language system in question. The
functions/semantics of an antipassive are nearly always *emergent*
properties which means that they cannot be identified by a single
'linguistic category' or so, but only by monitoring the complex behavior
of ALL (co-paradigmatisized) strategies, structures, and paradigms
involved in the process of forming antipassives (as well as in the
formation of the 'underlying' 'active' structure (i.e., ERG). The same
holds for passives, which as you surely know can occur in co-existence
with antipassives in a language system....

Finally, Pat says (in the same posting):

> Sorry, Wolfgang, after so much with which I can agree, I am unconvinced. I
> do not believe the essence of accusativity is in word-order.

I don't think that this is a matter of belief! It's just the question
how we are used to define ACC and ERG. If we claim that ACC and ERG
represent some kind of paradigmatic *behavior* that is scaled along the
Accusative Ergative Continuum then we should expect the AEC and its
instantiation(s) to present in ANY language, even in isolating ones. If
you look at English (not my best language, as all of you will know), we
HAVE to decide whether the pair

(5)	The woman saw the dog.
(6)	The woman went downtown.

follow an ACC or an ERG strategy. Because A in (5) ('the woman')
*behaves* like S in (6) ('the woman'), we can say that there is soem
kind of S=A behavior. If you look at other language systems with a more
or less fixed word order you will easily see that word order is crucial
for the AEC. The same naturally holds for languages that have OTHER
(morphosyntactic) means to react on the prerogatives of the AEC....


[Note: My email address has been modified: Please use
W.Schulze at!]
| 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

[ Moderator's transcription for those who have problems with 8-bit text: ]

(4)	Bi dund surguule/ t{\"o}gs-{\"o}{\"o}d end ir-sen
	1Sg:NOM middle schol:ACC:indef finish-CV:PERF PROX come-VN:PRET
	'After having fished the middle school, I came here.'

More information about the Indo-european mailing list