Re: Ergativity Question
| From: | Thomas R. Wier <trwier@...> |
|---|
| Date: | Wednesday, August 11, 2004, 5:48 |
|---|
From: Caleb Hines <cph9fa@...>
> > > Granted, 4) is a bit akward in English, and requires the use of a
> > > prepositional phrase, but its essentially the same as 2) but with a
> > > reversal of word order and voice.
> >
> > That's actually not true. In a truly transitive sentence, both
> > arguments *must* be present. In this case, the passive in English
> > allows an *optional* agent in an oblique phrase.
>
> I guess when I said "essentially the same" I meant semantically, not
> grammatically. The bottom line is that both 2) and 4) mean that there is an
> object called 'corn', an object called 'me' and and action called 'to eat'
> which the 'me'-object applies to the 'corn'-object. This brings me to...
Right. No one's disputing that passives and actives are truth
conditionally the same (unless you bring in quantifiers, which we
won't). But your question was about a grammatical phenomenon, so
I answered on that assumption. Besides which, semantics is *part*
of grammar.
> > You're still not getting at the question of how transitivity is
> > encoded in your language. That's absolutely crucial:
>
> The short answer is: it isn't encoded -- at least not in the grammar.
I think what you mean by "grammar" here is really morphosyntax. As
I said, semantic properties of languages *are* grammatical, and are
quite distinct from pragmatics, so it's still not clear to me how your
discussion shows that the language is ergative.
> Akathanu is very much an "engelang" (it was originally inspired in a
> predicate logic class), and some of the design goals include a sort of
> orthogonality and minimalism. One result of this is that transitivity is not
> marked in the grammar. In a sense, every argument of a 'verb'* is an
> optional argument -- no arguments are required. If the semantics of a word
> require some argument for a it to make sense, and there isn't a
> corresponding oblique word in the sentence, a default word ('something')
> is understood.
Ah. Well, then you'll still have a problem, because in all human languages
known to me, even if a verb is not syntactically transitive, it may still
be semantically transitive. The system you describe doesn't sound very
realistic, but more importantly for our purposes, it would make impossible
any kind of ergativity. The reason is that ergativity is a *structural*
relation (whether in syntax, morphology, semantics, etc.) between the
single argument of a one-place verb, and the patient argument of a two-place
verb. But your system sounds like it's abolishing the notion of argumenthood
entirely, thus making it impossible to encode a relation between any two
of those arguments.
[skip grammatical exposition]
> So, although it may not be strictly correct in grammatical terms to think
> of it this way, I think of the "patient-of-" and "agent-of-" markers as
> denoting a "passive voice" and "active voice" respectively. In my original
> question, I was thinking of 'me' in 5 as being 'ergative' because it is a
> non-subject agent, and 'corn' in 6 as being 'accusative' because it is a
> non-subject patient. My case markers might end up being more like indicators
> of semantic role, than grammatical case.
> Well, maybe. What morphological case the agent gets put in does not
> necessarily indicate that the agent has typical subject properties.
> That's the crux of the issue.
I'm not quite sure what you mean. Could you clarify or give an example?
> > This is not a good example to use in English, since "eat" is a
> > labile verb (neither strictly transitive, nor intransitive). In
> > a true antipassive, there is usually some morphological indication
> > that the patient has been demoted and the verb is detransitivezed.
[...]
> Labile? Now there's a term they don't teach in Language Arts 101! But I
> can't seem to think of any verbs that are only transitive. Could you give a
> good example, so I can see if my non-transitive system can handle it?
Sure. There are thousands of them. Here's a good minimal triplet:
Purely intransitive: dine. "I dined", *"I dined the food"
Labile: eat "I ate" "I ate the food"
Purely transitive: devour *"I devoured", "I devoured the food"
> > Actually, this is still not an ergative system. What you describe
> > here is simply a special instrumental case. Many, many nom/acc
> > languages have something like this. Russian, e.g.
>
> Well, the ergativity only came about by wanting a non-oblique case for a
> non-subject agent. I guess what I'm actually questioning is voice.
But you have still now shown that it is in fact an ergative construction!
> > Active voice and antipassive voice are not at all the same. Antipassive
> > and passive voice forms of verbs (or probably more accurately: related
> > verbs found in the lexicon) are *detransitivized* forms of active
> > transitive verbs. The only difference between them is which argument
> > gets demoted -- that associated with the agent-role, or that with the
> > patient-role.
>
> I think now I can see how we might be describing two sides of (almost) the
> same coin. While you define alternate voices as a means of _demoting_
> arguments from being required to being optional, I was thinking of them as a
> means of _promoting_ arguments, whether optional or not, to the (required)
> position of subject. So if I'm understanding you correctly, then with an
> originally transitive verb:
>
> Active Voice: Agent (required), Patient (required)
> Passive Voice: Agent (optional, oblique), Patient (required)
> Antipassive Voice: Agent (required), Patient (optional, oblique)
>
> Is that correct?
More or less, yes.
> Since in English, the subject of a sentence is always required, and with an
> active voice transitive verb it is typically the agent, I was (apparrently
> incorrectly) viewing voices like this (bear in mind that I wasn't thinking
> about transitivity at all):
>
> Active voice: Subject = Agent(required), Object = Patient(may or
> may not be present)
> Passive voice: Subject = Passive(required), Object = Agent(may or
> may not be present)
> Antipassive voice: The oppsosite of passive, and thus the same as active.
I am by no means an expert on Phillipine-style languages, but
I think this is definitely the wrong analysis for English. With the
exception of a few lexically defined labile verbs, transitivity
is a real feature of English, and active transitive verbs simply
must have both agent and patient arguments. Passives differ
in that the agent argument does not have typical subject properties
(various kinds of binding relations, agreement features, etc.).
AFAICT, English has no real antipassive, but if it did, the agent
would have such features, and the patient would not have object
properties.
> I would say that's a vast oversimplification of the Tagalog facts,
> based on what I've read and heard. Topicality and focus are central
> to Tagalog grammatical relations encoding. [...]
>
> Actually, I stumbled upon this description of Tagalog, which has some
> elements similar to what I was trying to describe:
>
http://www.atoni.com/dila/languages/LanguageFacts/Tagalog.htm
>
> What it calls "Objective voice" and "Agentive Voice" match what I was
> previously calling "Passive Voice" and "Active Voice".
Reply