The authors present results of a generalizability study of running record assessment. They conducted 2 decision studies to ascertain the number of raters and passages necessary to obtain a reliable estimate of a student's reading ability on the basis of a running record assessment. Ten teachers completed running record assessments of 10 first-grade students on 2 leveled reading passages. Findings indicate that each student assessed with running records should read a minimum of 3 passages to produce a reliable score. Using a fully crossed design in which all students were rated by each rater on all passages did not provide sizable advantages over a nested design in which students were crossed with passages and nested in raters. (Contains 2 tables and 7 figures.)