accusative and ergative languages

Stefan Georg georg at rullet.leidenuniv.nl
Mon Jun 21 13:21:00 UTC 1999


(> = P.R.)

>Thank you for your interesting (to me, at least, and I do not mean that
>sarcastically in the slightest) response/

I may add that I myself enjoy this exchange with someone who doesn't take
my occasional snide personally  (nor do I : polemike: me:te:r panto:n).

>Ralf-Stefan, because of your background, you are obviously going to be able
>to leaf through _Tipologija Jazykov Aktivnogo Stroja_  or _Printsipy
>Kontensivnoi Tipologij_  much more felicitously than I. My Russian is
>rudimentary, and though I can struggle through a passage, you can much more
>easily pinpoint what you may disagree with specifically. If you feel Klimov
>has erred grievously, let me know where and I will try my best to defend him
>(if I agree with that particular point).

While it is true that I do read Russian with ease, (and I did read at least
the first title years ago; and it is true that Klimov has interesting and
well-informed things to say on the various alignment-phenomena in the
languages of the world; add his Ocherk obshchej teorii ergativnosti) I
could do this job for you, but I feel disinclined to do so here and now.
Reason # 1: I don't see why I should look for arguments defending a
position which I find unattractive for a host of reasons, most of them
having to do with an empirical-based general rejection of the idea of
stadialism.

The second reason is that it seems to me that you hope that Klimov advances
reasons and arguments unbeknownst to you, which Larry and I won't be able
to deal with no matter how hard we tried, to win the day. No way.
But you say that you came to Klimov's conclusions independently, so the one
thing we should discuss here is your line of argumentation, for this is the
only thing which matters.

>As far what I agree with, my website makes very clear that I subscribe to
>Klimov's idea of a progression in development in language through
>neutral-active-class-ergative-nominative types. Klimov believes, and I
>agree, that this progression is *necessary* ab origine.

There's a lot to disagree with here. Whether there is really a "neutral"
type of alignment, seems very doubtful to me (note that this would be a
language where A, S, and O [I assume that you are familiar with this
convention; if not, see Dixon's publications on ergativity] are treated
*always* *alike* in every respect. Neutral case-marking is OK, neutral
cross-referencing properties OK, but some subsystem of such a language, and
be it word-order, will always give away a definite alignment, ERG, or ACT).
A "class" type is equally dubious, noun classification being a
morphological technique which can do a lot of things (inter alia, it can
enshrine accusative or ergative alignment, to be sure), but as such, a
morphological technique, it stands outside of the core issue.
The general idea of stadiality is, as I said, doubtful in itself, to say
the least, moreover, as I and now Larry have repeated several times, there
is no sense in the overall label of "ergative language", this being only a
(sometimes, if properly understood) useful impressionistic designations for
languages which show ergative features somehow "salient" for the average
European eye (e.g. ergativity by case-marking, where the whole thing was
first detected and named).

>Pat responds:

>I evade nothing. And why I should feel it incumbent on me to defend every
>last jot and tittle of Klimov's views perfectly escapes me.

The major lines of argumentation would do ;-)

>I have told you
>above where I agree with him.

Yes, you agree with the bottomline ("stadiality is a fact"). But why ?

>But you cannot have it both ways. You have consistently implied that some of
>my views are so far from the mainstream that, a priori, they must be wrong.
>This takes the insufferable form of "linguists agree that ..." as if my
>views may be equated with those of the fishmonger you mentioned above and no
>linguist would hold them. Klimov is one linguist who does hold views that I
>share, and this effectively debunks the notion (on this idea anyhow) that it
>is somehow intellectually disreputable to believe that certain laguage types
>grow naturally out of other language types.

This passage is hard to understand. For the record: I don't say that any
views, just by virtue of being far from the mainstream, must a priori be
wrong. The mainstream can be wrong (and is so very often, e.g. on the
illusion that there is something like the Altaic family of languages, just
to place my favourite running gag).

My fishmonger-example wants to express that it doesn't matter who utters a
view, but only the arguments do.

And, the other way round: a position is, imho, not automatically immune to
criticism just because it is held by otherwise well-reputed figures. They
can do wrong. The roster of eminent scholars, asserting that Turkic and
Mongolian are genetically related languages, is impressive and deservedly
so. Yet they are wrong.

>> That Sumerian is just another split-ergative language, since you seem to
>> doubt my (oh, not *my*, I could refer you to *eminent* linguists ;-)

>Pat, aside:

>And why do you not if such exist?

Because I don't have to, being able to defend my points on my own. If you
want a reading-list on ergativity, I could give you one, of course.

>Pat answers:

>Since you are writing of what Sumerian is or is not, I would think you might
>have understood that the point of my remark was that, although th markings
>of the maru: may be fairly well established, there is absolutely no
>agreement on what grammatical role these endings signify.

There is hardly any overall agreement about anything in linguistics, given
that a lot of journals in the field still accept anything they are handed
over. If there is disagreement on this particular point on your side, state
it and give your reasons.

>Pat responds:

>Yes, you may have dropped something here. To make this simple, why not give
>me your definition of "imperfective aspect", and I will attempt to find a
>maru: sentence that may be interpreted non-imperfectively. Personally, I
>believe *most* maru: indicate a progressive nuance rather than imperfective
>aspect.

No contradiction here. No, I'm not going to expose "my" definition of
aspect here (but you may with profit read, e.g., Comrie's handy book on the
issue). Only so much: if a language displays a contrast in aspect, *and*
possesses a systematic means of coding progressivity, then the progressive
forms will either coincide with those of the imperfective aspect, *or*
build upon them (i.e. they will belong, morphologically speaking, to the
"imperfective" system of that language, never to the perfective system).
Progressivity is functionally incompatible with perfectivity.

(OK, for clarity's sake: the gist of "my" definition of aspect is that,
while perfective aspect describes an action as an unstructured whole,
imperfective aspect draws attention to its internal structure, i.e. having
or not having beginning and end, filling a certain stretch of time, being
divisable into phases, being the pragmatic background of a narration aso.
for a subset of the notions which are most often associated with
imperfective aspect as against perfective aspect).

Note that this is not germane to my argument on an ergativity split in
Sumerian, whether or not the interpretation of maru:-HamTu as aspect-coding
inflections will hold water is not my issue here, nor do I regard myself
competent enough to decide this issue. There is a split, and is along the
line of *some* TAM-category-distinction. That's enough. The aspect-show is
in a different theatre.

>> lugal-le Hi-li ib2-dim2-me. "The king fashioned the wig"
>> The king is case-marekd as ERG (-(l)e), and the wig is ABS, so, in terms of
>> case-marking a perfect ERG construction.
>> The verb form, here given in transliteration, is morphologically to be
>> analyzed: i- (conjugation prefix for maru: or imperfective (horribile
>> dictu),

>Pat interrupts:

>This is certainly not the interpretation of *any* Sumerologist, linguist or
>philologist, of whose ideas I am aware. Where did you get it?  No one says
>i- is a conjugation prefix for maru: (or imperfective!) unless you got this
>from Gonzalo (?).

I see no reason to expose the scattered sources I'm using at the moment,
for this will inevitably lead to a rather tiring exchange of the type: "Oh,
that shoddy book, no wonder you found that drivel there".
What  *would* be useful though, would be if you named the sources you
trust, so that I can confine my search for instances proving my point -
that Sumerian is a split-ergative language, like any other one, thus
removing one cornerstone of stadialism - to these. No doubt I'll find them
there as well.

>The consensus view is, indeed, that -b- in this position is supposed to
>reference an inanimate patient. Why you might think that lugal, which means
>'king', and is probably as agentive as any noun can be, should be referenced
>by -b- totally escapes me. If anything, it would be referenced by -n-, which
>is, in the consensus view, connected with animates.

This is correct. Actually I said so a few lines above (now deleted to easy
our moderator's task), but failed to pay attention a few minutes later. The
center holds, though, which is about things being treated alike and things
being treated differently.

>Then you write: "we should expect the *patient* ... ".  What typlogist has
>enunciated this doctrine? Why should the animate subject of an intransitive
>(two-element) sentence be expected to be marked the same as the *inanimate*
>patient of a transitive (three-element) sentence???

I have certainly overlooked the animacy-category here. So, I happily accept
your correction that it should be cross-referenced by -n-. But it isn't
(see below).

>Or do you consider yourself a typologist?

With your kind permission, I do consider myself a typologist.

>The key fact that you pass by
>unremarked is that the subject of the second construction is in the
>absolutive; its ending is -0. Therefore, your "case-marking is of course not
>ergative" is simply incorrect. The case marking is ergative, which calls for
>the agent of an intransitive verb and the patient of a transitive verb to be
>marked with -0.

Only a slightly sloppy formulation on my part this time. I should have said
"S is marked by the absolutive case here, not by the ergative, as we are
entitled to expect". The overall case-marking structure of this
intransitive sentence is, of course "ergative", which included the
"ergative-case" being precluded here. In German, I would have used
"Ergativ" for the case and "ergativisch" for the construction.

> Additionally, the consensus view is that du is an irregular
>maru: of g[~]en so it needs no special maru: inflection.

Again, this is correct and a welcome correction, but not germane to the issue.

>Another point is that one group of Sumerologists considers various vowel to
>represent oral as well as nasal articulations derived from -n. i{3}-du
>*could*, according to them, represent *i{3}(n)-du.

If this is correct, this could eventually force me to admit (no, not that
Sumerian is not a split-ergative language, it is) that my chosen example
was not unambiguous enough to drive my point home (since the scribe *could*
have intended his from to be read /indu/).

Let's, then, look at further examples. I)f you don't mind, I'll help myself
to the ones you supplied yourself a few lines down in connection with the
pronouns:

>g[~]a-e i-ku{4}-re-en, 'I entered'

>(Yes, the -e here is another problem.)

>the -en cross-references the intransitive subject.

>In the transitive sentence, the -en is supposed to cross-reference the
>agent:

>g[~]a-e sag[~] ib-zi-zi-en, 'I am raising (my) head'.

I coul d be arrogant and say: sapienti sat. Sumerian shows accusative
features. But I think I'll explain ;-):

- en in both cases, the intransitive and the transitive one,
cross-references what we latinate accusative-minded Europeans usually call
a "subject"; or, iow, it cross-references *both* the agent of the
intransitive, as well as that of the transitive sentence above. That's
pure-vanilla accusativity, and nothing else. For, under ergative
conditions, both constituents are kept apart, by whatever means, yet here
they aren't, they are  treated alike in terms of verbal cross-referencing,
and that is what accusativity is all about. Get the message ?
What it exactly is, which conditions this split in Sumerian may be
discussed elsewhere (for I think we have tried the patience of this IE list
enough with Sumerian, especially since I am a tiro in this field, however,
tiro or not, I usually can tell  an instance of split-ergativity when I see
one, and here I do see one), aspect or shmaspect, the point is only that
*some* split exists, as, I repeat, in *every* other "ERG-Language", the
reverse not being the case for ACC-Languages, now it's your turn.

>As might be obvious by the corrections I have made above to your "analysis"
>of these Sumerian constructions, I continue to question what you have
>demonstrated.

You have corrected mistakes which show my shallow standing in Sumerian, but
which concentrated on marginal issues only. Finally you presented me with
the one pair of examples which saved me from some hours of digging in
Sumerian boobs (though this would have been educational for me). You have
my thanks, and I have, of course, now your agreement.

>The very facts you have detailed above have led me (but not many
>Sumerologists) to question whether -n- and -b- are patient/agent
>cross-references. If there are not cross-references, then your argument has
>very flat feet.

Sure, but this is a definition-trick. All birds can fly. Ostrichs and
penguins can't, so let's rewrite our definition of birds as to exclude the
latter species and be happy with it.

>> It is true that g[~]a[2].e and za.e are formally ergative cases, by virtue
>> of -e. However, I'm unaware of a systematic contrast between ergative and
>> absolutive forms (i.e. without -e) used in a clear-cut ergative way in the
>> language.

>Pat responds:

>That can hardly be my fault.

No, it may be the fault of the extant data, but these are the thing we have
to live and work with.

The same examples cited above, by means of which you demonstrated the
ergative-split in terms of cross-referencing on the verb do at the same
time show that, this time in terms of overt case-marking, the personal
pronouns do not discriminate between transitive and intransitive agents,
both being expressed by the same form (which may or may not represet
*originally and formally* an ergative form in -e).

>And what sources are those? And why would g[~]a.e develop into g[~]a when we
>see this no where else where -e is employed. Your sources have seriously
>misinformed you.

No, they haven't. Everything which you had to correct in my representation
of what I found, can be learned from the grammars I consulted. I take full
responsibility for everthing I misread or interpreted too rashly instead of
reading a few lines/pages down, as I should have.

My sources are, i.a., Hayes, Thomsen, Diakonoff, Poebel, Deimel,
representing different generations of Sumerologists with different
backgrounds and linguistic standing, certainly also with different
standings as connoisseurs of the language. However, if you dislike one or
more names on this list, feel free to say so, but if the only Sumerian
grammar you trust is that which you have written yourself, please share it,
instead of complaining on others. The best grammar is, here as elsewhere,
the text corpus itself, and I think the examples discussed so far, as long
as they don't turn out to be  entirely cooked up, make it perfectly clear
that Sumerian is no exception to our general knowledge of "ergative
languages". You will have to come up with a different one, which should not
be too difficult, since you are reconstructing the mother of all languages
;-)

>Your first problem is that you want to make a distinction (transitivity and
>intransitivity) for Sumerian verbs that is not really appropriate; these are
>not categories of the Sumerian verb in a real sense.

Again you caught me red-handed using sloppy and careless language. Of
course, transitivity is first and foremost a feature applicable to
*constructions* (clauses, sentences), and not of verbs. Mea culpa, but
again nothing to do with the fact that Sumerian is a split-ergative
language.

>Pat responds:

>I have already expressed myself on "pure" anythings. In the case of
>Sumerian, however, you have not demonstrated any accusative features.

That may be right, for, strictly speaking, you have. Chapeau.

Stefan



More information about the Indo-european mailing list