Abstract: A knowledge representation scheme for A expert systems-the Associative Network Based on Rules (ANBR) is proposed in this paper. B y this scheme, the knowledge acquired by an intelligent system is represented in two forms: the outer and the inner. The former is a knowledge structure based on rules and used to acquire knowledge and manage the knowledge base, while the latter is a partitioned associative network and used to direct inference control. Because the ANBR combines the understandability and modularity of production rules and the, powers of knowledge indexing and associative inference in networks, it becomes an effective scheme for knowledge representation.