The reliability of visual scoring of alpha electroencephalogram activi
ty in nonrapid eye movement sleep was evaluated. Three trained scorers
independently, on two occasions, assessed one record from each of IO
patients. In addition to standard sleep staging of the records, alpha
activity was scored using three methods: a) while the record was being
scared, an assessment of alpha activity during the preceding stage wa
s made at each stage change; b) after the record was scored a single g
lobal rating of alpha was made; and c) each 20-second epoch of the rec
ord was rated on the amount of alpha present. Relative and absolute ge
neralizability coefficients in excess of 0.810 and 0.698, respectively
, were obtained. However, the results also indicated that careful atte
ntion to methodological factors is critical in ensuring satisfactory s
tandards of reliability.