TURKISH CLAUSE LINKAGE^543
(15) Ben çahş-madan önce git-ti.
I work-MEDEN before go-ρτ
"He left before I worked."
(16) Ben cevab-ι bil-meden önce o (cevab-i)
I answer-ACC know-mEdEn before PRO
bil-iyor-du.
knOW-PROG-PT
"Before I knew the answer, he knew (the answer)."
(17) *Ben cevab-i bil-meden (cevab-i) bil-iyor-du.
I answer-ACC know-mEdEn PRO know-PROG-ρτ
Note that in (15) and (16), as expected in clausal juncture, there is no
coreferentiality requirement between arguments of the clauses. The sen
tence in (17) is ungrammatical because there is no postposition marking the
embedding of the dependent clause. This is permitted only when a -mEdEn
clause occurs in core juncture:
(18) Kitab-i oku-madan cevab-i bil-iyor-du.
book-ACC read-mEdEn answer-ACC know-PROG-ρτ
"Without reading the book, he knew the answer."
(18') Ben o-na söyle-meden cevab-i bil-iyor-du.
I PRO-DAT tell-mEdEn answer-Acc know-PROG-ρτ
"Without my telling him, he knew the answer."
Unlike the earlier examples, these forms involve no postpositions but have
an interesting restriction: either the subject or the object (indirect object
here) must be shared by both clauses, suggesting that it is a nonsubordinate
nexus at the core level. That is, when the clauses occur without a postposi
tion, a tighter linkage is involved. Though the core junctures to be discus
sed below require coreferentiality between subjects, this is an example of
what F&VV considers a "less restrictive" core juncture in that the corefer
ence may be either actor-actor or undergoer-actor (1984:194). If the post-
positionless constructions involving -mEdEn are truly instances of core
juncture, we would expect the clausal operators to have scope over both
clauses. This is, in fact, the case as the following example shows:
(19) Kitap oku-yabil(*-miş)-meden her soru-nun cevabini
book read-ABIL-mIş-mEdEn every question-GEN answer