Advances in Role and Reference Grammar

TURKISH CLAUSE LINKAGE^537

Second, they claim that all basic juncture types involve one of these
three layers. Thus they give examples of clausal juncture, in which full
clauses are joined; core juncture, in which cores are joined and thus neces­
sarily share clausal operators; and nuclear juncture involving the joining of
two nuclei which share all core and clausal arguments and operators.
Third, they claim that instances of clause linkage traditionally consid­
ered to be either coordination or subordination actually manifest three (and
possibly four) types of nexus. Coordinate elements manifest no dependency
at the level of juncture and no embedding of one within another. Subordi­
nation always involves one element embedded within another, specifically
functioning as an argument of the other. Cosubordination differs from
coordination in that it manifests a dependency relationship between the two
elements at the level of juncture. On the other hand, unlike subordination,
cosubordination does not involve the embedding of one element in the
other. There is a "grammatical category dependency" (Van Valin 1984)
rather than just a distributional dependency. For example, in core cosubor­
dination one core, though not an argument of the other will share the core
operator of modality with the other core. In the analysis of Turkish clause
linkage below, the subordinate nexus types generally (though not always)
correspond to what have traditionally been called "verbal nouns" in English
grammars of Turkish, while the nonsubordinate nexus types (instances of
coordination and cosubordination) often correspond to what have been cal­
led "gerunds".

2. On Turkish operators

Since operators and their assignment to particular layers of the hierarchy

play such a crucial role in the claims of F&W, it's necessary to discuss what

types of operators occur in Turkish. One of the most fascinating operators

in Turkish is -mI§. Probably the most insightful discussion of this suffix is

found in Slobin and Aksu (1981). They show that it functions as a marker

of aspect, tense and/or evidentiality. Interestingly, they claim that the his­

torical development of its function followed the order aspect —» tense —>

evidential (their "modality") and report further that developmental studies

show the same order of acquisition among children learning Turkish. This

matches precisely the prediction of F&VV that "the basic tendency in dia-

chronic developments [and acquisition] for clause operators is for more-