Theiling Online    Sitemap    Conlang Mailing List HQ   

THEORY: Referent Tracking (Was: (Re: THEORY: Information Structure; Topic/Comment, Focus/Background, Given/New.))

From:Chris Bates <chris.maths_student@...>
Date:Wednesday, November 23, 2005, 8:24
Since we're talking about Topics I'd like to share something that I
posted on the ZBB a while ago (and sadly got few replies). It's
indirectly relevant to the issue of Topics (or at least one of the
purposes of Topics). Here it is:

"I've been giving a lot of thought to referent tracking recently, since
I started making deeper forays into the area of syntax books. There seem
to be a few major ways that languages use to keep track of referents in
discourse and resolve ambiguity:

Basically works by indicating if some argument of the verb (usually the
subject) is the same as an argument of the preceding or following
clause. Usually switch reference systems don't have a fully functional
voice system for promoting other arguments to the priledged role
(subject) since if they did it would kind of defeat the purpose of the

This is what one of the syntax books I read calls a system with a
reasonably well developed voice system (like English) that is used to
keep track of referents via processes like zero anaphora. Eg in English,
if you have two consecutive clauses with the same subject you can delete
the second:

the man went to the store and 0 bought a coke

and you can use the voice system to maintain subject continuity:

the man went to the store but 0 was hit by a bus on the way

Although such voice changes can be used simply to delete arguments, they
are also widely used in English and other languages to maintain subject
continuity as a form of referent tracking. This is actually one of their
primary functions that is often overlooked, which is why languages with
an alternate means of reference tracking often lack a well developed
voice system.

*Topic* systems, while not obviously about reference tracking, seem to
play a major role in the tracking of referents in some languages such as
Japanese and Mandarin, where topics control zero anaphora. The choice of
a *topic* is often a fairly discourse stable referent which is then
assumed to be an argument of the verb where the number of arguments
present does not match the valency (transitive, intransitive etc).

In many Amerindian languages especially there is a system very much like
a *topic* system in its function except for the fact that verbal
morphology plays a large role (unlike extremely *topic* based languages
I know, which typically lack verbal agreement). Basically one argument
in a stretch of discourse is designated proximate, all others are
obviate, and then verbal agreement and other markers keep track of what
role the proximate plays in the discourse. The choice of proximate seems
to be a little more constrained than the choice of *topic* in *topic*
languages though.

Some languages, especially those like Swahili and the other Bantu
languages with a large number of genders, mainly use their extensive
agreement systems to keep track of referents.

Those are the obvious systems for reference tracking, but there are
probably more.

This area doesn't seem to have been giving much thought or through
describing by people writing conlangs, which is why it interests me. I
mean, people say "I have a switch reference system" or "I have a *topic*
system" etc, but they don't go into detail when it comes to the role
such systems play in things like reference tracking, and indeed whether
for a given system that is one of its functions (Swahili, for instance,
has a passive but does not use it as a major reference tracking device).
So your thoughts? I'd love to hear details about how various natlangs
accomplish reference tracking functions, and also about how any of your
conlangs do it."

Although obviously I missed off discussion of things like pronouns, my
main focus at the time was on ways that most languages track a core
argument through a dialogue or story, rather than how they track all
arguments, especially since often knowing the identity of one argument
helps you work out the rest.


Rik Roots <rik@...>Referent Tracking