Abstract
Studies in educational psychology in Japan are reviewed from the viewpoint of reliability. Cronbach’s alpha is one of the most frequently used —if not fully understood— measures of reliability. The main purpose of this note is to give an overview of what alpha is and is not. Specifically, it is shown that (1) alpha is the mean of all split-half reliabilities, (2) alpha is a lower bound of reliability, and (3) alpha equals reliability under essential tau-equivallence. It is also shown that (1) alpha is not a measure of unidimentionality or homogeneity, (2) there are criticisms on understanding alpha as a measure of internal consistency, (3) alpha depends on a variety of characteristics including number of items, and (4) alpha is not a greatest lower bound. Alternative measures of reliability and future issues are discussed.
