Declarative models of syntax

Theoretical linguistics—which from its earliest days has self-consciously modeled its working methodology on that of the natural sciences—continues to face a dilemma of long standing which interestingly does not seem to be an issue in the established empirical disciplines. On the one hand, parallel to the situation in other scientific domains, the investigation of the conditions characterizing well-formed human language is logically independent of the question of how a procedure can be implemented that produces or licenses grammatical utterances. Yet, at the same time, early generative syntax and more recent research in the wake of Chomsky (1995) does not separate the meaning of a theory from its procedural, computational realization. Instead, the theory itself is directly tied to the mechanism by which a given string is procedurally derived. With linguistic generalizations directly tied to procedural mechanisms, progress in generative linguistics arguably has been hampered by ordering paradoxes and other theory-internal issues, and the procedural mechanisms or metaphors have stood in the way of more empirically adequate, declarative generalizations (cf. section 2)—deficiencies which have led to the development of declarative linguistic frameworks. Declarative approaches to syntax, such as Head-Driven Phrase Structure Grammar (HPSG, Pollard and Sag, 1987), Lexical-Functional Grammar (LFG, Bresnan, 1982) and others (cf., Borsley and Börjars, to appear), follow the otherwise standard scientific practice of seeking a declarative characterization of the research domain and its properties, i.e., they investigate the constraints which are observed to hold of linguistic representations. Declarative models of syntax are not committed to any particular processing model or procedural mechanisms for deriving linguistic representations.