This report summarizes computational investigations of language comprehension based on Marvin Minsky's theory of frames, a recent advance in artifical intelligence theories about the representation of knowledge. The investigations discussed explored frame theory as a basis for text comprehension by implementing models of the theory and developing methods for exploiting its power; specifically they: (1) investigated how common-sense knowledge about the world can be embedded within a frame system, how that knowledge can be made "active" to model the intentions of a person planning a simple event, and the extent to which knowledge embedded in one particular frame system can be used to understand a grammatical construction called "noun, noun modification"; (2) implemented a testbed system that understands connected discourses using knowledge embedded in frames to "glue together" the various pieces of information conveyed by individual sentences; and (3) studied how natural language text of very simple sorts can be immediately analyzed into the meaning of that text by exploiting the structure of individual frames. The first chapter of the report outlines frame theory, discusses briefly how this theory is embedded within Frame Representation Language (FRL--a programing language), and summarizes the major results of the investigations. The remaining three chapters discuss the results in detail. (FL)