Ergativity correlations

[Jon Aske]

> In Spanish, there are what I call "ergativity effects".  For example, the
> default order of intransitive clause subjects and transitive clause
> direct objects is postverbal.  It turns out that this is grounded in
> discourse: topics are preverbal, focus elements are postverbal.  It could

I'm not sure that I would call the Spanish phenomenon you describe
"ergativity effects".  Spanish intransitive subjects are more often than not
topics and not foci.  And when they are foci they can also be preverbal (F1
position) as well as postverbal (F2 position) or rheme-final (F3 position)
(the same thing applies to transitive subjects).  It is true that overt
intransitive subjects are about as often foci as they are topics and thus
there is a high percentage of overt intransitive subjects that are
postverbal, especially with some intransitive subject arguments with certain
semantic characteristics ("roles").

Ergative subjects can also be foci, just less frequently than absolutive
subjects.  This is for two reasons: 1) objects are typically better focus
candidates than (transitive) subjects; and 2) transitive propositions make
poor thetic (topicless) assertions.  Thus most transitive subjects that are
foci are "contrastive foci" and not "new foci".

I am sure that you are aware of this complexity, but I am afraid that
simplifying this picture by saying that Spanish displays "ergativity
effects" may lead to more confusion than illumination.

Going back to the original question, correlations between grammatical
categories and other properties of language can be found by looking at the
functional source of the grammatical categories and then seeing what those
functional properties correlate with and why.  In other words, we must look
at the non-grammatical source of grammatical categories in semantics and,
especially, information structure (discourse pragmatics).

As many have argued before, I believe that the category subject, or
nominative, is a grammaticalization of the ubiquitous informational category
topic.  Transitive and intransitive subjects are the default topics for a
predicate (most topical arguments in the abstract).  The category
absolutive, on the other hand is not as well motivated a category.  Overt
(full nominal) objects are often foci, but not so highly accessible ones
(unless they are contrastive).  And overt intransitive subjects are foci
about half of the time and topics the other half.  Then, of course, there
are the semantic affinities between O's and S's, most clearly evidenced in
causativization and anti-causativization constructions, for instance (e.g.
transitive vs. intransitive "boil"), in which the S of the non-causative
predicate is the O of the causative version.

So how and why do the absolutive and ergative categories arise?  It seems
that it happens by the reinterpretation of certain constructions in which a
"patient" acts as an S (e.g. passive constructions).  Such constructions are
then reanalyzed, presumably to take advantage of the newly acquired argument
coding morphology (cf. e.g. Myhill 1992, Typological discourse analysis).
This is probably also facilitated by the informational and semantic (weak)
correlations, or affinities, between S and O arguments.

>>From this picture we can ascertain, or begin to look for, correlations
between grammaticalized ergative and absolutive categories and other
characteristics of a language.  For instance, we may ask: what kind of
languages use passive constructions and in what contexts?  Is seems that
rigid word order languages often use passives, since languages with free
word order achieve the same effects by simply rearranging the topic and
focus arguments.  And passives are often used with perfective aspect, since
in such assertions the topicality of the patient is increased and thus
patients are more likely to be the topic of the assertion (note, however,
that passive can also be used in a language such as English to place a
patient in F1 (preverbal) focus position, whereas a more flexible order
language, such as Spanish would just place the object in F1 position without
changing its grammatical relation or the basic construction.

Anyway, I'm getting carried away again.  Sorry about that.  I just wanted to
clear up some potential misunderstandings.

Jon