The purpose of this study was to identify superintendent evaluation items which have the power to discriminate or measure differences among superintendent performance. Based on a review of literature, superintendent evaluation instruments, job descriptions, and critical work analyses, a questionnaire containing 87 criteria was developed and tested in the spring of 1987. Completed questionnaires were received from 451 participants consisting of board members, superintendents, administrators, and others in 30 school districts located in Iowa, Wyoming, Missouri, and Michigan. A minimum of 15 raters from each school district used a five-point Likert-type scale to rate performance of the superintendent in order to test the items on the questionnaire;Two pools of discriminating items were identified. The first pool of 71 items was to be used by groups of 15 or more raters. The second pool of 51 items was to be used by groups of seven or more raters. All the items contained in both pools of items have the power to discriminate based on a statistical model developed by Dr. Menne and Dr. Tolsma. The model, which is a variation of an analysis of variance (ANOVA) test, required three factors present to assure performance was measured: (1) use of multiple raters, (2) raters within a group must closely agree, and (3) rating must indicate a difference among the individuals rated. A discriminating item was one which generated minimum variance among raters in a group and maximum variance among superintendents. A Cronbach alpha reliability coefficient was calculated for each of the two pools of items;Individual rater questionnaires were classified into district enrollment sizes, rater positions, and cohesion factors of the seven board members. An analysis of variance (ANOVA) was calculated on each of the discriminating items based on the various size, position, and cohesion classifications. Duncan's multiple range test was used to identify group means with significantly different values for each of the items;The performance items which have the power to discriminate among superintendent performance identified in this study should be used as a basis for board members and superintendents to develop a written superintendent performance evaluation instrument.