Many clinical evaluations are subjective, resulting in ordinal level measurements. A widely used example in medical rehabilitation is the Functional Independence Measure (FIM), which provides a measure of disability. The FIM is an 18-item, 7-point Likert scale ranging from complete dependence to complete independence. Parametric statistics are commonly used for the analysis of ordinal data. However, Likert scales often lead to violation of many underlying assumptions. This study examined the comparative power of the t test with the Wilcoxon rank-sum test using real pretest/posttest data sets measured on an ordinal scale. FIM scores were obtained on 714 geriatric patients at admit and discharge from a rehabilitation hospital. A Fortran 77 program was written to sample with replacement from each admit and discharge data distribution. Results indicated the Wilcoxon rank-sum test outperformed the t test for almost every sample size and alpha level examined.