This memorandum, part of a series, critically examines the fixed-effects analysis of variance procedure used in an intensive classroom study. The procedure, used by Gentile, Roden and Lein (Journal of Applied Behavioral Analysis, 1972, 5) for intensive studies of single subjects, is found inappropriate because the basic assumptions of an analysis of variance model are typically violated when continuous data on a single subject is gathered over time. Two other proposed analyses of variance models for single subjects are also considered and found unsuitable. Time series analysis, which takes serial correlation effects into account, and a median-based method are recommended as alternatives to analysis of variance designs. (Author/LRP)