Objective. To determine the National Institutes of Health Stroke Scale's (NIHSS's) association with upper extremity (UE) impairment and functional outcomes. Design. Secondary, retrospective analysis of randomized controlled trial data. Setting. Not applicable. Participants. 146 subjects with stable, chronic stroke-induced hemiparesis. Intervention. The NIHSS, the UE Fugl-Meyer (FM), and the Arm Motor Ability Test (AMAT) were administered prior to their participation in a multicenter randomized controlled trial. Main Outcome Measures. The NIHSS, FM, and AMAT. Results. The association between the NIHSS and UE impairment was statistically significant (P = -0.204; p = 0.014) but explained less than 4% of the variance among UE FM scores. The association between NIHSS total score and function as measured by the AMAT was not statistically significant (P = -0.141; p = 0.089). Subjects scoring a "zero" on the NIHSS exhibited discernible UE motor deficits and varied scores on the UE FM and AMAT. Conclusion. While being used in stroke trials, the NIHSS may have limited ability to discriminate between treatment responses, even when only a relatively narrow array of impairment levels exists among patients. Given these findings, NIHSS use should be restricted to acute stroke studies and clinical settings with the goal of reporting stroke severity.