In this paper fuzzy rule inconsistency resolution and fuzzy rule insertion methods are proposed for fuzzy neural networks. Necessity support and possibility support (referred to as support pair) are applied to detect and remove inconsistencies. In addition to the support pair, the concept of initial learning point is used to handle rule insertion. We demonstrate the use of the proposed methods in an example called the Knowledge Base Evaluator (KBE). After inconsistency resolution operations, learning is improved. Moreover, a new fuzzy rule is generated by setting initial learning point based on deleted conflict rule. The result of using rule insertion is much better than with inconsistency resolution alone.