accusative and ergative languages

[Patrick C. Ryan]

Dear Wolfgang and IEists:

 ----- Original Message -----
From: Wolfgang Schulze <W.Schulze at lrz.uni-muenchen.de>
Sent: Tuesday, July 20, 1999 5:15 AM

Wolfgang wrote:

> 16. Do YOU use a special theoretical frame work in order to substantiate
> your claims? If yes, which one?

Pat responds:

This is an excellent suggestion --- theoretically --- but I do not think it
is going to happen on this list. Answering your pertinent questions as they
should be answered would involve writing a book at least the size of
Dixon's. But, good thought!

Wolfgang continued:

> Now, let me finally turn to some claims by Pat:

> Pat wrote:

> Wolfgang wrote:

>>> Hence, there are NO "ergative (or "accusative") languages" or
>>> only, if you use this term in a very informal sense.

>> Pat wrotes:

>> In the sense you are using these, they seem to be of little value in
>> describing anything.

Wolfgang responded:

> I can only repeat what I have said in my earlier postings: ACC and ERG
> are nothing but the structural (and structured) reaction to more general
> principles of language based information processing. The polycentric
> architecture of language systems that encode these principles allows
> that the individual centers of this polycentric cluster react
> differently on these principles. Here, I cannot elaborate the underlying
> frame work which is labeled "Grammar of Scenes and Sceanrios" (GSS) and
> documented in Schulze 1998, chapter IV, but let me briefly say that I
> propose a (more or less) universal cause-effect 'vector' (C->E) itself
> metaphorized from underlying figure-ground relations (F->G) to be one of
> the most dominant principles of information processing. This vector can
> be weighted which leads to a continnum (here shortened) C->0(e) C->e >
> C->E > c->E > 0(c)->E (capitals represent heavy domains, small letters
> represent light domains). An ACC strategy would be to behave in a C->e
> sense, an ERG strategy would infer c->E. Note that these vector
> representations are NOT (in themselves) language specific or sensitive
> for specuial (linguistic) categories! Different categories may BEHAVE
> differently with respect to this continuum, regardless their own
> architectural make-up. ANYTHING in a language system may be sensitive
> for the AEC (accusative ergative continuum) as long as it is relevant
> for encoding the cause-effect vector (and its derivations). Hence, CASE
> may play a role just as AGR, word order, subject assignment,
> topicalization, discourse cohesion, co/subordination, paradigmatization
> of speech act participants and much more. But they may play their roles
> DIFFERENTLY! The description of their roles heavily depends from the
> diachrony of the paradigm in question, its formal architecture as well
> as its integration in a co-paradigmatic context ('structural coupling'
> in a broader sense). From this it follows that the individual centers of
> ALL language systems have to react upon the universal demands of the
> C->E vector, disregarding their internal architecture. Only IF ALL
> relevant centers (and they are many, I grant!) behave in one direction,
> THEN we are allowed to call the language system (or better, its
> Operating System) ACC or ERG. However, it is a much more difficult task
> to describe intermediate states that allow ACC in some parts of the
> Operating System, and ERG in some others. Here, we have to establish a
> (motivated!) hierarchy first of co-paradigmatic structures (such as CASE
> and AGR, CASE and word order, AGR and Personality, AGR and Noun Classes,
> to name only some). In such structures, one part sometimes is more
> dominant than the other with respect to its behavior on the AEC. If we
> can describe such dominant behavior we can refer to the whole structure
> as either more ACC or ERG. In a second step we have to go on describing
> the higher levels of this hierarchy (which itself should find an
> adequate linguistic explanation based on an appropriate language
> theory). Finally (and ideally) we would arrive at a term that would
> describe the functional dominance of one of the poles on the AEC with
> respect to an Operating System (not a language system) in toto. Only
> then, and I stress, only THEN we are allowed to use the term ergative or
> accusative with respect to an Operating System (for which 'ergative' or
> 'accusative language' would be an informal label).

Pat responds:

It would take my background than I have to properly apperciate your argument
above but, from what I *can* understand of it, I concur. As I wrote in
another post, and I hope I understand the above well enough to validly
connect it, I believe the critical factor is how cause-effect is mapped but
that the underlying cause-effect is the reality behind the masks.

>> Wolfgang wrote further:

>> Pat wrote:

>> I am sorry that I do not agree with the validity of this distinction (actant
>> vs. agent). For me, 'actant' is 'agent'. Perhaps you can explain the
>> difference.

Wolfgang explained:

> In terms of Functinal Grammars (as well as in GSS) 'actants' refer to
> ALL such linguistic expressions that encode a referential entity in a
> clause. Hence, in a sentence such as 'I met John several times in
> Chicago', 'I', 'John', 'times', and 'Chicago' are (abstract) actants
> that play different roles in the scene. But only 'I' is a linguistic
> agent, whereas John plays the role of a patient etc. Note that 'agent'
> and 'patient' are labels for semantic hyperroles (or macroroles in the
> sense of Foley/VanValin). I think that such a distinction is very
> helpfull. It is based on strong theoretical arguments and helps to avoid
> many false or at least problematic generalizations. A much more
> controversial (and much more difficult question is to define the labels
> 'subjective' (S), 'agentive' (A), and 'objective' (O) which should not
> be immediately equated to neither 'subject'/'object' nor to
> 'agent'/'patient'. S, A, and O are highly abstract terms that describe
> more structural than semantic or syntatic properties.

Pat responds:

Although I do not care much for 'actant' and think it has little to
recommend it over, say, NP, it is, of course, the prerogative of theorists
to introduce and define new terms as they choose.

Otherwise, if I understand you properly, I concur again.

>> Wolfgang wrote:

>>> The fact, however, is that many 'ergative' languages lack an
>>> antipassive. For instance, there are nearly 30 East Caucasian languages
>>> all of them using some ergative strategies in at least parts of their
>>> operating systems. But only a handfull of them (five or six, to be
>>> precise) have true antipassives (only one has some kind of
>>> "pseudo-passive").

>> Pat wrote:

>> Are you asserting that the majority of ergative languages do not have
>> anti-passives?

Wolfgang answered:

> I do not assert anything in the sense of 'ALL language have...'. Even
> the claim 'the majority of ergative languages (sic!) have...' is rather
> suspect to me. What I said is that in those 'ERG systems' I looked at
> (about 200) antipassives are rather the exception than the norm.

Pat responds:

I am not surprised really. I suspect the (original [?]) real function of an
anti-passive is suggesting lessened effective agency. What do you think?

>> Pat wrote:

>> As for the far-reaching conclusions of Dixon based on discourse cohesion
>> strategies, in my opinion they are flawed because these strategies are
>> purely conventional.

Wolfgang answered:

> Language systems ARE conventional! This is one of the major points of
> language tradition and L1 acquisition (despite of minimalism etc.).
> Discourse probably is one of the most important factors in the
> emergence, organization, and dynamics of language systems. We should not
> refer to the abstract notion of context-free 'sentences' that would be
> responsible for for grammatical 'events'. Such a view stems from the
> tradition of Classical Philosophy which is an INTERPRETATION of what
> goes on language. Today, we have become used to think of language in
> single sentences, to brak them up the way we do etc. But this is an
> analytic tradition, not part of the ontology of language itself, which
> is much more synthetic in nature than we are used to think. - A sentence
> does not function but in its co-text (as well as in its con-text). All
> sentence internal strategies used to be embedded in the techniques of
> co(n)textualization. No wonder, that ACC and ERG also work in this
> direction (though they may appear as more 'autonomous',
> sentence-internal mechanisms secondarily, especially if a language
> system as developed separate means to indicate discourse cohesion).

Pat responds:

Again, I agree. A valuable caveat, and often neglected.

 <snip>

>> Pat wrote (on Lak, East Caucasian):

>> I have to plead ignorance of Lak however, I find an analysis of an otherwise
>> completely ergative sentence as having an ACC word-order impossible. I
>> believe this opinion rests on a false analysis of word-order significance.

Wolfgang wrote:

> I tried to show that in Lak only some sentence patterns show 'complete
> ergativity'. Most of them show a mixed paradigmatic organization with
> respect to morphology (CASE, AGR). WOrd order is another VERY imporant
> aspect of the AEC. Consider e.g. a language that has canonical

> #SV
> #AOV

> (# = sentence boundary). Here, S behaves like A with respect to #, hence
> it is ACC. In #SV vs. #OAV it is S=O which indicates ERG behavior. The
> problem of actance serialization is also addressed in polypersonal
> systems (without CASE, e.g, Abxaz, Lakotha etc.). It IS a very important
> indicator for ACC, ERG, no question. Anything else would refer to the
> linguistic tradition of say 60 years ago (Bloomfieldian tradition).

Pat responds:

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.

>> Wolfgang continued (on Lak):

>>> But if you say "I am surely hitting you (plural)", you get:

>>> na b-at-la-ti-s:a-ra zu
>>> I:ABS I:PL-hit1-DUR-hit2-ASS-SAP:SG you:PL:ABS

>>> Here we have:

>>> ACC with respect to word order
>>> ERG with respect to class agreement
>>> ACC (or neutral) with respect to case marking)
>>> ACC with respect to SAP agreement (-ra is triggered by 'I:ABS').

>> Pat wrote:

>> So, agent marked ABS and patient marked ABS is, according to you, ACC with
>> respect to case marking? Sorry, not convincing at all! And I think the SAP
>> difference can be explained as indicating seriality of the patients involved
>> as recipients of the action by *one* agent.

Wolfgang responded:

> I said ACC (or neutral!) with respect to case marking! That means that A
> lacks ERG marking (which is OK with Silverstein). Semantically, it means
> that S is encoded like A (which is ACC). That fact that O is NOT in an
> accusative-like case does not argue against this assumption, because
> CASE is here structrually coupled with ADGR which allows us to classify
> AGR as ACC *here*. The SAP difference in Lak cannot be explained in the
> sense Pat proposes, Lak totally lacks such strategies.

Pat responds:

Sorry again. Perhaps it is the term "accusative" that is preventing me from
grasping your argument.

>> Wolfgang wrote further:

>>> Now, please tell me: Is Lak an 'ergative' or an 'accusative' language?
>>> [Please note that I did not include (among others) strategies of
>>> discourse cohesion, reflexivization and logophization].

>> Pat wrote:

>> By these examples, I would tell you that Lak is 'ergative'; and that
>> accusativity is not demonstrable from these examples.

Wolfgang responded:

> If you refer to the standard (morphological), but rather obsolete
> interpretation of ERG you may be right (but not for SAP NPs). But,
> fortunately, the ACC/ERG typology has freed itself from such a narrow
> interpretation of ACC/ERG which is nothing but a very small excerpt from
> the over-all typology.

Pat writes:

"Unchain my heart, and let me be free, dum-dee-dum . . .  (:-{})

>> Pat wrote previously:

>>>> I think it is likelier that, because of perceived greater animacy (or
>>>> definiteness), pronouns have a different method of marking that can still
>>>> be interpreted within an ergative context.

>> Wolfgang wrote:

>>> This a (very simplified) 'on-dit' that stems from the earlier version of
>>> the Silverstsein hierachy. Again, we have to deal with the question,
>>> whether a 'pronoun' (I guess you mean some kind of 'personal pronouns')
>>> can behave 'ergatively' or 'accusatively'. The list below gives you a
>>> selection of SAP case marking in East Caucasian languages with respect
>>> to ABS/ERG:

>>> ABS vs. ERG ABS = ERG
>>> ALL ---
>>> Singular Plural
>>> Plural Singular
>>> 1.Incl. Rest
>>> 1:SG Rest
>>> 2:SG Rest
>>> 1:SG/PL Rest
>>> --- ALL

>>> This list (aspects of personal agreement NOT included!) shows that SAP
>>> pronouns may behave differently within the same paradigm. Any
>>> generalization like that  one quoted above does not help to convey for
>>> these data...

>> Pat wrote:

>> I do not have a reference book for Lak so that my hands are somewhat tied.
>> But, I have found that paradigms are often inconsistent in ways that reflect
>> earlier lost phonological changes, or other lost schemata. The locative
>> plural terminations of IE are certainly not, in origin, terminations of the
>> locative plural. Etc.

Wolfgang responded:

> Nothing the like! The list I gave refers to what can be described for
> East Caucasian languages in toto! Most of these paradigms are
> functionally motivated, I grant (schulze in PKK 2 (Schulze 1999) will
> tell you the whole story).

Pat concludes:

Still not convinced on this point . . . and, I am trying hard.

Pat

PATRICK C. RYAN (501) 227-9947; FAX/DATA (501)312-9947 9115 W. 34th St.
Little Rock, AR 72204-4441 USA WEBPAGES:
http://www.geocities.com/Athens/Forum/2803 and PROTO-RELIGION:
http://www.geocities.com/Athens/Forum/2803/proto-religion/indexR.html "Veit
ek, at ek hekk, vindga meipi, nftr allar nmu, geiri undapr . . . a ~eim
meipi er mangi veit hvers hann af rstum renn." (Havamal 138)