This paper argues that the standard "When it is expected that a test will be used to make norm-referenced assessments of groups rather than individuals, normative data based on appropriate group statistics should be provided," which was considered secondary in the 1985 "Standards for Educational and Psychological Testing" and essential in the 1974 standards, should be considered conditional; that is, primary for some situations and secondary for others. When tests are to report building averages, it is better to use school or group norms rather than pupil norms. To support this assumption, two sets of data from standardized tests are compared: (1) data for pupil norms and school average norms for grades 2, 6, and 8 obtained from published norms tables for the Iowa Tests of Basic Skills, 1978; and (2) data for pupil norms and "small group" and "large group" norms obtained from the norms tables for the Stanford Achievement Tests, 1982. The results show that median pupil scores and the median school average scores are seldom the same. Illustrations show that distribution of school averages is considerably less variable than the distribution of pupil scores. A copy of the class average performance and its relationship to school average performance for a university economics course is attached. (JAZ)