Anomalies such as redundant, contradictory, or deficient knowledge in a knowledge base indicate possible errors. Various methods for detecting such anomalies have been introduced, analyzed, and applied in the past years, but they usually deal with rule‐based systems. So far, little attention has been paid to the verification and validation of more complex representations, such as nonmonotonic knowledge bases, although there are good reasons to expect that these technologies will be increasingly used in practical applications. This article does a step towards the verification of knowledge bases which include defaults by providing a theoretical foundation of correctness concepts and a classification of possible anomalies. It also points out how existing verification methods may be applied to detect some anomalies in nonmonotonic knowledge bases, and discusses methods of avoiding potential inconsistencies (in the context of default reasoning inconsistency means nonexistence of extensions).