Our natural language comprehension system, “Sparser” , uses a semantic grammar in conjunction with a domain model that defines the categories and already-known individuals that can be expected in the sublanguages we are studying, the most significant of which to date has been articles from the Wall Street Journal’s “Who’s News” column. In this paper we describe the systematic use of default syntactic rules in this grammar: an alternative set of labels on consitutents that are used to capture generalities in the semantic interpretation of constructions like the verbal auxiliaries or many adverbials. Syntactic rules form the basis of a set of schemas in a Tree Adjoining Grammar that are used as templates from which to create the primary, semantically labeled rules of the grammar as part of defining the categories in the domain models. This design permits the semantic grammar to be developed on a linguistically principled basis since all the rules must conform to syntactically sound patterns.