Error classification methods are used throughout aviation to help understand and mitigate the causes of human error. However, many assumptions underlying error classification remain untested. For example, error is taken to mean different things, even within individual methods, and a close mapping is uncritically presumed between the quantity measured (errors)and the quality managed (safety). Further, error classifications can deepen investigative biases by merely relabeling error rather than explaining it. This article reviews such assumptions and proposes alternative solutions.