Local and Long Distance Reflexives in Turkish
Jaklin Kornfilt
Syracuse University
The purpose of this paper is to study the
core facts of Turkish reflexives. (Only morphologically unbound forms
will be studied here.) It will be shown that these facts can be
adequately explained by teasing apart factors concerning q-theory, morphology,
binding theory, and discourse factors (e.g. topic-orientation, point
of view etc.).
A number of previous studies have suggested
that Turkish has no genuine reflexive elements, but rather only a
kind of logophoric element whose distribution is entirely determined
by discourse considerations. While this is perhaps true for a
written, literary style, it is not for the standard-oral as well as
written-styles, or at least not for a majority dialect. In that
dialect, first and second person reflexives are always locally bound.
The third person reflexive has two forms: one is inflected (for
person and number), and the other is non-inflected. The non-inflected
form is a local anaphor and thus must be bound by a c-commanding
local antecedent. (We find some dialect variation with respect to
subjects of embedded domains; for some speakers, such subjects can be
bound from outside their own domain-but within the subjacent domain-,
if they are co-indexed with weak AGR.)There are certain well-defined
exceptions to the c-command condition; this paper will address those
exceptions and propose independently motivated solutions (two of
those are: possessor raising in unaccusative contexts, and
reconstruction of scrambled subjects into their original positions).
Further, some constructions will be discussed that are problematic
for Binding Theory, in the sense that the complementary distribution
predicted for anaphors versus pronominals doesn't obtain. It will be
shown that those instances are limited to adjuncts, in particular to
those expressed as postpositional phrases, thus further motivating a
q-theoretical
approach as advocated by a number of researchers for other
languages.
For the majority dialect that distinguishes
between locally bound versus long distance-bound reflexives, the
element used as the second type is morphologically more complex, as
mentioned earlier. This is the opposite of what has been observed for
a variety of languages. I argue that the Turkish long distance
reflexives are not problematic for that cross-linguistic
generalization, for two reasons. The first is that the Turkish long
distance reflexives are (a special kind of) logophors rather than
genuine long distance relfexives; their distribution is determined by
discourse-related factors rather than structural ones of the type
found in long distance anaphors in other languages (e.g. sensitivity
to intervening blockers with different j-features). Since such
logophoric elements do not need to be c-commanded, they can't be
analyzed as bound variables, either. The second reason why the
morphological complexity of the (apparently) long distance-bound
reflexives in Turkish don't pose a problem for the (apparently)
universal generalization about complex anaphors as locally bound
elements is my analysis of the Turkish forms as inflected heads of
DPs, whose specifier is pro. It is that
pro,
I claim here, which is the logophor, while the morphologically
complex anaphor is indeed locally bound, namely within the DP by its
pro-specifier. If this analysis is right, Turkish
reflexives have properties that are diametrically opposed to those
imputed to them by some researchers: they obey strictly local,
structural (and thematic) conditions. It is only where their
inflection licences an associated pronominal that they appear to be
able to "reach further" for an antecedent. This analysis, which is
based on viewing pro as a logophor in certain constructions, is also
consistent with the observation that overt pronouns, wherever they
occur in syntactic contexts where pro is also
licensed, function as elements disjoint from their potential
antecedent. In such situations, the overt pronoun is not ruled out by
Condition B, but rather by the availability of pro for the
coreferential reading, which then results in the apparent behavior of
pro
(and, where it occurs, of the inflected reflexive linked to that
pro)
as a bound variable.