Objective: To develop a reliable and valid test of cognitive function suitable for patients with severe cognitive impairment. Design: Administration of a test; test-retest reliability; comparison to traditional test. Setting: Chronic long-term-care facility. Patients or Other Participants: The participants were 40 elderly residents with severe cognitive impairment. Main Outcome Measures: Results of the Test for Severe Impairment (TSI) and its subsections (language, memory, executive function, and motor performance); correlation of test and retest scores; correlation of TSI with Mini-Mental State Exam. Results: The TSI was significantly correlated with the Mini-Mental State Exam (r = 0.83, P less-than-or-equal-to 0.0001). Test-retest reliability was high (r = 0.96, P < 0.0001). The internal reliability of the test was also good (alpha = 0.90). Preliminary result of a factor analysis suggests that factor scores may be derived that relate to memory, language production, and knowledge of body parts. Conclusions: The TSI is a valid and reliable test of cognitive function in patients with severe cognitive impairment. It is appropriate to use it as a unified scale.