In this paper we consider the merging of the language of feature structures with a formal logical language, and how the semantic definition of the resulting language can be used in parsing. For the logical language we use the language EL, defined and implemented earlier for computational semantic purposes. To this language we add the basic constructions and operations of feature structures. The extended language we refer to as ‘Generalized EL’, or ‘GEL’. The semantics of EL, and that of its extension GEL, is defined model-theoretically: for each construction of the language, a recursive rule describes how its value can be computed from the values of its constituents. Since GEL talks not only about semantic objects and their relations but also about syntactic concepts, GEL models are nonstandard in containing both kinds of entities. Whereas phrase-structure rules are traditionally viewed procedurally, as recipes for building phrases, and a rule in the parsing-as-deduction is viewed declaratively, as a proposition which is true when the conditions for building the phrase are satisfied, a rule in GEL is best viewed as a proposition in Dynamic Semantics: it can be evaluated recursively, and evaluates not to true or false, but to the minimal change in the model, needed to make the proposition true. The viability of this idea has been demonstrated by a proof-of-concept implementation for DPSG chart parsing and an emulation of HPSG parsing in the STUF environment.