[...] somewhat different approach to fitting together morphology and syntax is suggested by work in Relational Grammar (e.g. [Davies, W., 1981, "Choctaw Clause Structure", UCSD, San Diego], [Harris, A., 1981, "Georgian Syntax: A Case Study in Relational Grammar", Cambridge University Press]). In this framework, it is emphasized that processes of grammar can be sensitive to grammatical functions (or relations) not only at a surface level or at an underlying, quasi-semantic level, but at any number of intermediate levels as well. In fact, it has been suggested that the distribution of a particular morpheme can be characterized by predicates that include variables that range over all levels of description. Thus, first a "relational network" for a sentence is constructed, which is basically a global representation of a complete syntactic derivation, including explicit instances of GF-rules such as Passive. Then a morphological form for the verb in question is constructed, based potentially on properties of this relational net as a whole.
- Baker (2004), a pag.35-36