anaphors vs. logophors

[John Moore]

I'd like to comment on John Myhill's claim that no formal account of the
distribution of jibun is possible.  Since I don't know anything about
jibun, I will address a question that John didn't raise - namely whether it
is desirable to have a formal account of reflexive pronouns in English.  I
suspect that the answer to this question bears on John's original one, but
I'll leave that to people who know something about jibun to decide.

One of the big problems with the classic GB (LGB - Chomsky 1981) binding
theory was that it attempted to provide a unified account of the
distribution of `himself' in (i-ii):

(i) Patrick admires himself.
(ii) Pictures of himself embarrass Patrick.

What (i) seems to illustrate is that reflexives require local antecedents -
this can be stated quite simply in any of a number of ways, using any of a
number of formal devices (e.g., in terms of tree-theoretic concepts like
c-command, argument structure terms, grammatical relations, etc.).  (ii),
on the other hand, is problematic for these straightforward formal
approaches, and led Chomsky to introduce a host of complications, including
a disjunctive definition of SUBJECT and the notion of accessibility, which
were not independently motivated.  Even with these devices, there were
still problematic data - e.g.,

(iii) Physicists like yourself are a godsend.

It seems that the basic flaw was trying to unify uses of reflexives that
shouldn't have been unified - the desire for global parsimony often comes
at the expense of local parsimony and leads to greater abstraction.

Building on previous work by Cantrall, Kuno, and others, a number of formal
accounts emerged in the early 1990s - including Reinhart and Reuland (P&P),
Pollard and Sag (HPSG), and Dalrymple (1990).  These accounts distinguished
uses of reflexives that could be easily treated in a sentence-internal,
syntactic approach, from logophoric uses that required a discourse-based
approach.  This simplified the formal syntactic approach considerably, as
its domain of application was reduced.  Furthermore, these approaches did
provide a means to tease apart the two uses - that is, it was not the case
that recalcitrant data was simply thrown to `some theory of discourse'.
Thus, by abandoning a unified account, one gains in local parsimony.

However, there is another possibility - and I suspect this may be what John
was getting at.  Given that some uses of reflexives *require* a discourse
account, could one develop a unfied discourse-based analysis on all uses of
English reflexives.  Will such an analysis suffer from the kind of global
parsimony problems that the LGB binding theory did?

The one proposal I know about is in an excellent paper by Zribi-Hertz
(Language 1989).  She raises these questions very clearly and argues that
the putative syntactic effects in reflexive binding are essentially
manifestations of the more general discourse-based effects found
independently in their logophoric uses.  In the end, however, her analysis
becomes rather complex, involves a number of disjunctive conditions, and
eventually has to admit that there are basic differences between
clause-bounded anaphora and long-distance bounded logophora.  Trying to
achieve global parsimony, again, seems to have a cost in local parsimony.

This is not to say that Zribi-Hertz's approach is not the right one - it
could be that all of the uses of English reflexives have a functional,
discourse explanation; it could also be that the logophoric uses have
discourse explanation, while the clause-bounded cases have a formal
account, albeit with a functional motivation, or it could be that there are
discourse-based and formal accounts that are essentially unrelated.  I
suspect that one's choice among these options will have a lot to do with
one's ideological beliefs about language.  I don't think it is safe to say
that a formal analysis is impossible for these facts - plently of
formalists appreciate the need for discourse factors as explanatory devices.

In the end, however, it seems that getting the right split between global
and local parsimony is part of what makes linguistics an art.

John Moore

http://ling.ucsd.edu/~moore/