The aim of this retrospective cohort study was to examine the accuracy of the El Escorial electromyographic criteria for the diagnosis of amyotrophic lateral sclerosis (ALS). Based on a consecutive series of 73 patients with suspected ALS, the sensitivity of electromyography (EMG) ranged from 0.2 in the cranial segment to 0.74 in the thoracic segment. Specificity was highest (0.92-1.0) in the cranial and thoracic segments. Using receiver operating characteristic (ROC) curve analysis to examine the effect of varying the number of abnormal segments required for the diagnosis, we found an overall accuracy of 0.90 and 0.87 if one or two abnormal segments, respectively, were required. The best combination of sensitivity and specificity was achieved by requiring EMG changes in two segments with abnormalities in a single muscle in the cranial and thoracic segments, but abnormalities in two muscles in the cervical and lumbosacral regions. These findings lend support to the EMG criteria proposed at El Escorial.