To analyse the psychometric properties of the EQ-5D in German stroke survivors undergoing neurological rehabilitation. The EQ-5D, the Hospital Anxiety and Depression Scale (HADS) and the Stroke Impact Scale (SIS) were completed before (210 subjects) and after (183 subjects) a patient education programme in seven rehabilitation clinics in Bavaria, Germany. A postal follow-up was conducted after 6 months. Acceptance, validity, reliability and responsiveness of the EQ-5D were tested. The SIS subscales were used as external anchors to classify the patients into change groups between the measurements. The proportion of missing answers ranged from 4.7 to 8.6%. Between 16 and 19% reported no problems in any EQ-5D dimension. At baseline, correlations between EQ-5D index and the SIS subscales ranged from 0.15 (communication) to 0.60 (mobility). Correlations with the EQ VAS were slightly smaller. All scores were reliable in test-retest with intraclass correlations ranging from 0.67 to 0.81. EQ-5D index and EQ VAS were consistently responsive only to improvements in health, showing small- to medium effect sizes (0.27-0.42). The EQ-5D has shown reasonable validity, reliability and, more limited, responsiveness in stroke patients with mild to moderate limitations of functional status, allowing it to be used in clinical trials in rehabilitation.