This dissertation presents a formal theory of Serbo-Croatian grammar. The theory predicts acceptable form/meaning pairs for a substantial chunk of Serbo-Croatian. In particular, we analyze Serbo-Croatian declarative and interrogative main clauses, embedded clauses, a couple of different types of nominal modification, control and predication, as well pro- and encliticization.
Linguistic expressions are represented as triples of typed terms, with each typed term modeling one of the following sets of properties of a linguistic sign: semantic (i.e. truth-conditional meaning), tectogrammatical (i.e. syntactic combinatorial properties), and, finally, phenogrammatical properties which specify the expression's linearization possibilities.
The focus of our work is on word order in Serbo-Croatian, which is very free in some respects but extremely rigid in others. With phenogrammar and tectogrammar as distinct components, we can isolate word order problems from tectogrammatical and semantic combination, and state theory-internal phenogrammatical generalizations. This is particularly important for the analysis of 2P enclitics, whose placement cannot be adequately characterized tectogrammatically.
The most elaborate component is phenogrammar. We postulate many different phenogrammatical types and modes of combination. This enables us to create islands of fixed word order, while still allowing free reordering of higher-level phenogrammatical objects. Of special significance are phenogrammatical terms which denote sets of strings. Such terms represent possible pronunciations of expressions which can be linearized multiple ways without a change in meaning. Essentially, we are modeling semantically insignificant reordering as phenogrammatical indeterminacy.
Our choice of grammatical architecture is empirically motivated, but methodological in nature. This dissertation purports to show that a decent theory of Serbo-Croatian word order can be given in a framework which does distinguish phenogrammar and tectogrammar; not that a comparable (or superior) theory cannot be given in one which does not. We do, however, believe that this project brings attention to languages with complex word order patterns and their role in delineating a realistic categorial grammar framework for linguistic analysis.