Case organization of verbs has provided a powerful mechanism for natural language analysis systems. However, only simple semantic-marker-like information has been used to determine the acceptibility of lexical elements as case-role fillers. Actually, this ability is influenced by more intricate relations among words. In addition, a case-based view of semantic knowledge often leads to the separate specification of each shade of meaning of a verb. These two problems are addressed in this thesis. A case-like organization of semantic knowledge which includes a network of relations among lexical elements is presented. Any piece of information contained in the system may be used as a case-frame specification, or it could be used as information which determines case-role fulfillment. Rules for the use of this information have been designed to permit a single case-frame to recognize many shades of meaning of a verb, even to the point of accepting metaphoric language use, The network of relations is hierarchically organized, and knowledge is retained at many levels of generalization. Along with the existence of case-organization in the network, these multiple levels provide some control over the traversal of the network. A small implementation is provided to demonstrate the use of a variety of strategies for fitting case-frames to input. The model is intended as a bottom-up component for the identification of those pieces of information which may be relevant to a given input.