This study compares four different item-analysis procedures with respect to their effectiveness in identifying items which would make up reliable and valid tests. The four procedures were: (1) the item-test biserial correlation. (2) the item-test covariance divided by the test standard deviation (or the item reliability index), (3) the "standard length" of the item, and (4) the ratio of the item validity index to the item reliability index. These procedures were applied to a 32-item, multiple-choice test with an essay test criterion. Results for a 300-case item-analysis sample and for three 100-case cross-validation samples are presented which show each of the four procedures to be about equally effective.
