Date: Fri, 1 Nov 1996 18:19:50 +0100 (MET)
From: Tanya Reinhart <Tanya.Reinhart@let.uva.nl>
To: Linguistics Conference <LINCONF@tamvm1.tamu.edu>
Subject: logophors
Message-Id: <Pine.SOL.3.91.961101181901.26828A-100000@cclsun01>
In-Reply-To: <Pine.SOL.3.95.961030154500.8115A-100000@suma3.reading.ac.uk>
Huang brought up an extremely important point in his comments on Tenny.
"A number of papers presented so far mention/discuss the notion
of logophoricity, which was introduced in studies of African
languages (I think) by Hagege. Naturally, most of these papers
refer (rather uncritcally) to the influential work by Reinhart
and Reuland. But I feel that RR's use of this notion is too
liberal. Whenever there is a counterexample to structural binding
theory, it is attributed to the factor of logophoricity. This
will eventually lead to making it impossible to permit the
recognition of any real counterexample and falsify any theory.
Unless the notion of logophricity can be clearly defined so that
we know what constitutes a clear set of counterexample to it,
the concept is of no great use (apart from the fact that it may
be used as an escaping route)."
I could not agree more with this theoretical statement (though I do
not think it actually holds for R&R's treatment of SELF anaphors).
There is, in fact, circularity in the way the term 'logophors' is
used in some of the anaphora studies.
One example is Sells' concept of 'pivot', used to account for the
distribution of 'zibun' in Japanese. Sells has three notions (which
he labels 'roles'):
a. Source = agent of communication
b. Self = person whose consciousness is reported
c. Pivot = Person whose perspective is represented.
The first two are easily recognizable. Often, there is a matrix verb
of communication or consciousness that signals this context. In
narrative texts where often point of view is not signalled in this
way there are always independent ways to identify the point of view
(temporal deixis, exclamations, absence of proper names, etc.) But
non of these could be found in Sells example (i)
i) Taro got wet, since Yoko poured water on zibun (zibun = Taro).
To rescue (i) Sells declares that it is the pivot role that enables
'zibun'. There is not independent way to know what it means for the
sentence to be from the perspective of Taro in this case. (All we
know is that the speaker says that Taro got wet, and Taro's
consciousness is not involved in this report. What else 'perspective'
could mean here, is not any clearer than saying that for some reason
or other, the speaker finds it more appropriate to use 'zibun' here,
rather than 'kare'.) 'Pivot', thus, is the escape hatch that Huang
mentioned, which enables the theory to handle any counter example,
and remain always true. (If e.g. we find a context where 'zibun' is
not allowed, we will just declare that it is not a pivot environment.)
Since the concept 'pivot' is undefined, it lends itself easily to
the construction of new theories. E.g. this is precisely the central
concept Tenny uses to justify a theory of why 'Max put a book near
him' is a logophoric context. Since, as Huang points out, by any
more explicit definition, it would not come out as such.
Another very common example of a circular use of the terms 'logophor'
is found in studies attempting to maintain the traditional BT condition
A for SELF anaphors. This condition allows (iia), since the anaphor
is bound in its GC, but it rules out (iib, c).
i a) Max likes pictures of himself
b) Mak thinks we should buy a picture of himself
c) The picture of himself that Max gave me was ugly. (based
on Jackendoff).
Much ingenuity has gone into showing that (iia) is an instance of
condition A, particularly in contexts of reconstruction and
intermediate SPEC positions. But then, of course there is the question
why (iib) is nevertheless good, in violation of condition A. Here
it is declared that the anaphor is in fact a 'logophor', so it need
not bother the standard condition A. If 'logophor' had a strict
definition, as proposed by Huang, we could immediately see that this
theory is wrong, since in (ic) there is no point of view, or any other
known definition of logophoric contexts, and still the anaphor is
allowed, in volition of BT condition A. However, once we also have
'pivot' in our inventory, we can easily declare that in (ic), the
anaphor is a pivot-logophor. Since this notion is undefined, BT
condition A always makes the right prediction: It selects an arbitrary
syntactic domain in which it allows anaphors to appear, and define
all counter examples as logophors.
This problem does not arise in R&R's treatment of SELF anaphors.
It was said already in R&R that the term 'logophor' in that framework
has no theoretical status, and no definition actually makes any use
of it. It is used mainly to make the connection with more traditional
studies of such anaphors. In fact, I think now it was a mistake to
use this term at all, in the discussion of SELF anaphors. As far
as reflexivity condition A goes there is only one relevant notion:
SELF anaphor (=an anaphor with the capacity to reflexive mark a
predicate). Condition A only says that when it occurs as an argument
of a (syntactic) predicate, this predicate must be reflexive. And
it does not say anything about all other occurrences, namely it allows
such anaphors to occur freely everywhere, accept as arguments of the
relevant type. In this respect, R&R's approach is the same as that
of Pollard and Sag.
In all sentences of (ii) reflexivity condition A equally allows the
anaphor. This is all that is meant by our saying that the anaphor
is a logophor in all three cases. Though I still suggest, in
retrospect, to avoid this usage of the term. It is very common in
all areas of binding that the binding conditions only rule out certain
occurrences of an anaphoric element. This means that as far as the
BT goes, all other occurrences are allowed. As a result, there are
many environments where more than one anaphoric option is allowed.
(In many PP contexts we can equally use a pronoun, a SE anaphor and
a SELF anaphor). There is an independent question regarding which
of the options allowed by the computational system will be preferred
in actual discourse. It is here that the concept of point of view
may be found relevant. As mentioned already in R&R, we do not believe,
however, that SELF anaphors in English are preferred over pronouns
only in point of view contexts. (We discuss point of view and focus
contexts, but nothing in the theory prevents their use for other
discourse functions, which we believe exist, but they are not relevant
for just now.)
One area which could appear problematic for R&R is where an anaphor
is permitted as an argument of a reflexive predicate ('This seat is
reserved only for myself'). Here we make an empirical claim, which
is certainly falsifiable, that this specific occurrence is possible
only when the anaphor is focus (stressed). Technically, we derive
this by assuming that in this case the focus-anaphor moves at LF,
hence condition A is not violated. We have, thus, a concrete
entailment that point of view contexts alone cannot license a SELF-
anaphor in that position. (E.g. Max believes that Lucie likes himself).
This is so since condition A will still rule it out, even if there
is very good discourse reason for it. An independent question which
I brought up in a previous comment (on Dravidian anaphora and Hamilton)
is whether the LF movement account of SELF-anaphors in focus context
is desirable on independent grounds, namely, whether this is the right
analysis of focus. I suggested an alternative line that could be
explored following Baker, which, as I explained, requires great care,
in view, of the question that Huang brought up, of how to avoid
circularity.
Note, finally, that the problem of logophoricity does arise acutely
in the area of SE (unlike SELF) anaphors (e.g. in accounting for the
distribution of 'zibun'). The standard assumption, e.g. on Icelandic
is that in the relevant local domain (down to infinitival clauses),
SE obeys c-command and subject orientation, but outside this domain,
where it is said to occur logophorically, it does not. There is a
problem of falsifiability in maintaining that the local context
generalization is indeed syntactically motivated. Reuland and
Sigurjonsdottir suggest that which binding by a syntactic process
(the feature checking of SE anaphors) always takes precedence over
processes of coreference. Hence, in the local domain where SE can
be syntactically bound this is obligatory, and it entails subject
orientation in the standard way. Only when this is impossible
coreference is free. (Reuland and Sigurjonsdottir "Binding in
Icelandic: Syntax or discourse", ms. 1996).
--Tanya Reinhart