Discusses learning roles and causal structures for capturing patterns and causality relationships. The authors present their approach for knowledge discovery from two specific medical databases. First, rules are learned to represent the interesting patterns of the data. Second, Bayesian networks are induced to act as causality relationship models among the attributes. The Bayesian network learning process is divided into two phases. In the first phase, a discretization policy is learned to discretize the continuous variables, and then Bayesian network structures are induced in the second phase. The authors employ advanced evolutionary algorithms such as generic genetic programming, evolutionary programming, and genetic algorithms to conduct the learning tasks. From the fracture database, they discovered knowledge about the patterns of child fractures. From the scoliosis database, they discovered knowledge about the classification of scoliosis. They also found unexpected rules that led to discovery of errors in the database. These results demonstrate that the knowledge discovery process can find interesting knowledge about the data, which can provide novel clinical knowledge as well as suggest refinements of the existing knowledge.