This work describes a procedure for correcting inconsistent knowledge bases. It resolves contradictions with as few changes as possible and is introduced for abstract logic based on a consequence operator. Elementary properties, such as the existence of a correction or the unchangeability of a consistent knowledge base during correction, are theoretically proven. Since a correction may be incomplete, I explain how it can be completed consistently. Since it is not unique in general, the selection of a correction is explained on the basis of additional information.