Non-accusative, non-ergative, non-active
|From:||Jim Grossmann <steven@...>|
|Date:||Friday, March 8, 2002, 22:35|
For anyone interested in this subject, I strongly recommend R.M.W. Dixon's
It seems to me that the easiest way to avoid being stuck with either an
accusative core grammar or an ergative one is to give each of these kinds of
arguments a different marking:
1) subject of intransitive verb
2) subject of transitive verb
3) object of transitive verb
May I back up and attempt a review of various types of core-grammars?
Correct me if I'm wrong, but IIRC...
Every language has a way of indicating that certain arguments stand for
referents in these relationships to the action denoted by the verb, and to
other entities where applicable:
S = the subject of the intransitive verb, the doer that is doing what the
verb denotes, but not to something else
A = the agent, the doer that is doing what the verb denotes to something
O = the object, which undergoes that which the verb denotes
In nominative/accustative languages, S and A are encoded in the same way,
with position before the verb in a declarative sentence, or nominative case,
which is most likely to be marked with a zero form. In English, in which
word order is used to encode S, A, and O, the fact that S and A both occupy
the same position in the sentence, and both dictate verb agreement, indicate
that English is a nominative/accusative language.
In an ergative/absolutive language, S and O are encoded in the same way.
Andreas was correct to point out that, in an ergative/absolutive version of
English, the word order would be VS for clauses with intransitive verbs.
However, in natlangs, IIRC, the S&O vs. A relationships are marked
morphologically, with absolutive and ergative cases respectively.
Absolutive is the case most likely to be marked with a zero form.
There are alternatives to these arrangements that occur in natlangs.
There are "split-ergative" languages, in which S & O are marked the same way
sometimes, and S & A are marked in the same way at other times, depending on
a number of possible factors, including the semantic properties of the noun
or verb. (See Dixon for details.)
There are "split-S" languages, in which S is marked one way for one class of
verbs, and in another way for another class of verbs. Usually, one verb
class stands for things that the S undergoes ("wilt," "be born," "fall") and
another class of verbs stands for things that the S controls or does
("write," "beat," "eat").
There are "fluid-S" languages, in which S is marked one way if the verb
stands for something S controls, and another way if the verb stands for
something S undergoes. Unlike verbs in "split-S" languages, a single verb
in a fluid-S language might take either marking, depending on the
relationship between the subject's referent and the action denoted by the
verb. So "The house burned" would be "Sobject V," but "The fire burned"
could be "Sagent V." "The man coughed" could be "Sobject V" for an
involuntary cough, but "Sagent V" for a voluntary cough. Verbs in a fluid S
language might translate differently into English according to the kinds of
subjects they have, so that "The woman-Sobject verbX" would be "The woman
fell," but "The woman-Sagent verbX" would be "The woman (deliberately)
dropped to the ground."
(Incidentally, my long-time project "Palo" is evolving into an attempt at a
WAIT, there's more! :-)
Although I heard somewhere that this is rare in natlangs, there's no reason
why you couldn't mark the "S" "A" and "O" arguments each in their own
distinct way, producing a non-accusative, non-ergative core grammar.
Still another kind of marking: semantic marking (as opposed to syntactic
IIRC, nominative/accusative & ergative/absolutive schemes fall under the
"syntactic" marking category, in which arguments are marked according to the
verb's syntactic properties, regardless of the semantic relationships
between the verb and its arguments.
For instance, in English, we can say "The man opened the box" and "The door
opened," placing both the word for "man" and the word for "door" in first
position (the same syntactic marking) even though "man" is the agent and
"door" is the patient (different semantic roles). Why? Because,
syntactically, "open" is an active verb in both sentences. To look at it
semantically, it *prototypically* stands for an action peformed by an agent.
In a version of English with direct marking, "the man" and "the door" would
be marked differently, something like this:
"The man-agent opened the box-patient." "The door-patient opened."
The marking on the arguments reflects the semantic relationships that the
arguments have with the verb as it is used in a given sentence, not
according to the verb's prototypical meaning.
* * *
The big picture is a lot messier and more interesting than I've made it look
here. Some of the schemes I've mentioned can be combined, and influenced by
semantics in many more ways than I can name. Ergativity has been such a
popular subject on this list, I can't recommend Dixon's book strongly enough
For those among us whose want to create the most exotic grammar possible, I
think that semantic marking offers the hottest lead. For instance, most
languages mark experiencers the same way they mark agents. e.g. Compare
"The man saw me," and "The man hit me." Mark stimuli and agents in the same
way, and you've got something that you may find pleasingly odd.
Semantic marking schemes need not be exhaustive or confusing: You could do
it with just two markers for agents vs. non-agents if you wanted to.
* * *
By the way, I've frequently been confused by the use of the term "active"
language on this list. Could anyone define "active language" for me?
Also, I'm not sure that I understand core-grammars that don't have much of a
transitive vs. intransitive distinction: could these be languages in which
virtually all the arguments are introduced by particles or adpositions of