ergativity arising

[John Myhill]

I am vaguely developing a theory about ergativity and word order based
on observations from Austronesian languages (though I'm aware of the
(correct) typological observations about ergativity being associated with
verb-final and verb-initial order. As I think most of you functionalists
know,
the Austronesian languages as a group sort of hover between ergativity and
accusativity, with some more in one direction, others more in the other
direction, and others who knows where in between. Basically, they (all?)
have some kind of construction with ergative/passive morphology (e.g.
direct object like intransitive subject, transitive subject different but
occurring
much more frequently than, e.g. the English passive. In some Austronesian
languages, this `ergative-passive' construction is statistically preponderant,
and these languages are popularly considered `ergative' (e.g. Samoan, and
I think it's acceptable to analyze Philippine languages like Tagalog this
way these days, e.g. Cooreman Fox & Givon in Studies in Language 1988 I
think it
was), while in others the `ergative-passive' construction is much less common.
It turns out that in Javanese, the ergative-passive construction (with the
prefix di-) occurs about 25% of the time (see my article in SIL), while in
Indonesian the cognate construction occurs about 40% of the time, while
in Cebuano the `goal focus' ergative-passive construction occurs about 50%
of the time and in Tagalog it's about 67% of the time (I think I'm
remembering
these numbers about right--Matt Shibatani reported the last two, if I
remember correctly). This correlates with the degree of word-order
flexibility in these
languages. Javanese is the most strongly Actor-Verb-Patient (SVO if you want to
use these terms, but they aren't appropriate)--the `ergative-passive'
construction in Javanese HAS TO be Patient-Verb, like a more normal
English-type passive construction. On the other hand, Indonesian is more
flexible--the Patient in the ergative-passive construction can go either
before or after
the verb (and is usually after), and it is comparatively more common than in
Javanese. On the other hand, Indonesian CANNOT normally have both the Agent
AND the Patient after the verb in this construction (actually I have found
a single
example of this, but this is out of several hundred tokens)--so Indonesian
word order is still restricted. On the other hand, Tagalog and Cebuano
freely allow
Verb-Agent-Patient or Verb-Patient-Agent constructions with their
ergative-passive construction, which is correspondingly much more common.
So, what this
means is that as we go from the verb-medial Javanese to the generally
verb-initial Philippine languages, the frequency of the ergative-passive
construction goes up and up and the language gets `more ergative.' For
example, there are di-constructions in Indonesian with postverbal patients
which could not be translated into Javanese with the di-construction but
rather has to use the `active' construction. Here's a clear pair:

        INDONESIAN
        Dengan  sigap   disambarnya     kayu ela...
        with    quick   seize   yardstick
        `Quickly, he seized a yardstick...'
(Suman 1978:161)

        JAVANESE
        Gage wae        deweke  ndudut  kacune...
        quickly she     pull-out        her-handkerchief
        `Quickly, she pulled out her handkerchief...'
(Brata 1979:167-8)

In both languages, the di- construction is normally used for temporally
sequenced transitive verbs (I think Paul Hopper first made this observation
about Indonesian), and in both languages the Patient can only go before the
verb if it has some special discourse salience. But since Javanese does not
allow postverbal Patients in the di- construction, this means that if the
Patient does not have any special discourse salience, the di- construction
cannot be used, even if the clause is temporally sequenced--neither
`kayu ela' nor `kayune' have any special discourse salience and so must
be postverbal, but this does not stop Indonesian from using its di-
(passive-ergative) construction, which can take a postverbal patient, but
this DOES stop Javanese from using ITS di- construction, which CANNOT take
a postverbal patient. This clearly shows how word order is related to the
frequency of a passive-ergative construction. And I bet we could find
similar pairs in Indonesian as opposed to Tagalog, where the passive-ergative
construction can only be used in Tagalog because it has even more word order
flexibility than Indonesian. This suggests that ergativity arises (in some
situations) as a verb-medial language gets more flexible verb-initial
order,
or, correspondingly, an ergative VSO/VOS language can becomes
nominative-accusative as it switches to more rigid SVO order.   John Myhill