Explicit representation of relations plays some role in virtually all higher cognitive processes, but relational knowledge has seldom been investigated systematically. This paper considers how relational knowledge is involved in some tasks that have been important to cognitive development, including transitivity, the balance scale, classification hierarchies, and deductive reasoning. It is proposed that important properties of these tasks are due to the fact that they entail processing relations. The nature of relational knowledge, its implementation in neural nets, and its role in higher cognition are examined. The paper outlines the properties of relational knowledge and distinguishes it from more primitive processes such as traditional associationism. Relational knowledge is described as flexible, explicit, and can be organized into complex structures such as lists, trees, and propositional networks. Relational knowledge is central to mechanisms that are basic to human reasoning, such as analogy and planning. The properties of relational knowledge are obtained at the cost of higher processing loads. Empirical criteria for relational knowledge are also indicated. (Author/KDFB)