The Functional Independence Measure (FIM) is one of the most widely used measures in the rehabilitation community. It has been used to monitor patient progress over time and to assess outcomes of rehabilitation. Measurement of change, however, has a problem. The change scores are interpreted differently on the measurement occasions due to unstable frame of reference (Wright, 1996). This could be true in FIM measure. The purpose of the study was to test the stability of the FIM items and to compare uncorrected measures with the corrected measures by Wright's correction method (1996). A total of 188 stroke patients' records were collected from the National Medical Center in Korea. The data were analyzed by a Rasch Rating Scale Model with FACETS program. Model-data fit was determined by Infit and Outfit statistics (<0.5 or >1.5). The standardized differences (Z) were computed to determine whether differences between admission and discharge estimates are valid indicators of change over time. Two FIM items, i.e., bladder and bowel movement, with poor fit statistics were eliminated from the final estimations. The rating scale steps were different across two occasions with greater than 2.0 of Z statistics from four of the six scales. Eleven items have absolute differences greater than 2.0 of Z statistics across two occasions. The results showed that some of FIM items and rating scale steps were functioning differently across two occasions, so that the differences in items and patient measures obtained from separate analysis were not stable and interpretable. In order to compare the true changes in patient and item measures independent from the changes in rating scale steps or items functioning across the two occasions, Wright's correction procedure was used. One hundred six patients' functional independences were significantly changed across two occasions. The comparison of patient measures revealed that misfitting patients were slightly fewer in the corrected method (30.3%) than in the uncorrected method (32.2%). Eight items' difficulty levels were significantly changed. The comparison of item measures revealed that misfitting items were slightly larger in the corrected method (62.5%) than in the uncorrected method (61.1%). The different results were found from four items out of 16 in standardized differences between the corrected and uncorrected method. Wright's method could provide a valid basis for analyzing changes when rating-type scales in the discipline are used to evaluate the true changes in item difficulty and person's performance. Keyword(s): assessment, measurement/evaluation, older adult/aging issues