Experimental studies of human error detection have not made much progress since the 80s and 90s (e.g. Allwood, 1984; Rizzo et al., 1987; Sellen, 1994; Zapf et al., 1994). One of the reasons is because there is not an established methodological paradigm. A connection between interruption and error detection is proposed – when an error is detected and corrected, it is similar to handling an interruption because the original task has to be suspended and resumed later. The connection is based on Trafton et al.’s (2003) characterisation of interruption, which dissects interrupted activities into various measurable time-based components. A similar anatomy of the error detection process is proposed giving time-based measures such as detection lag, correction lag and resumption lag; error detection examples are discussed. The main contribution of the proposed characterisation is a methodological one – postulating a set of dependent measures for future systematic investigations of the error detection process.