Reliable monitoring of electroencephalographic (EEG) and electromyographic electroconvulsive therapy (ECT) seizure duration has become important as these assessments have become a routine part of the clinical practice of ECT. In this regard, accurate automated seizure duration determinations would be particularly valuable. As a result, the present study was performed to assess the reliability of available computer-automated determinations of seizure duration (Thymatron Model DGx ECT machine; Somatics, Inc.) and to explore the factors upon which such reliability as well as the determinations of experienced raters depend. We found that the experienced human raters had very high interrater reliability, significantly higher than either did with the automated Thymatron DGx ratings. In general, the reliability of all ratings declined in the context of artifact, poor postictal suppression, or an EEG seizure end point that was reached gradually. Reliability was also greater for continuation ECT as compared with the index course. The reliability of Thymatron DGx versus experienced human ratings was particularly sensitive to these factors, ranging from 0.68 when several of these factors were simultaneously present to 0.999 when all these factors were absent.