Overestimation or underestimation of functional capacity in community-dwelling older people with cognitive impairment was evaluated between the responses of subjects and family members (proxies) by cognitive function level. Out of all the residents aged 65 years and over living in Yoita town, Niigata Prefecture in 2000 (n = 1,673), 1,544 voluntarily participated in the interview survey held at community halls or at home (92.3% response). They underwent the Mini-Mental State Examination (MMSE) for assessment of cognitive function and answered questionnaires comprising socio-demographic, psychological, physical and medical, and social activity items (2000/11). According to the age of the subject and MMSE score, we defined cognitive decline (MMSE scores < 1 SD below age-specific means, n = 371). 158 pairs among 371 subjects with cognitive decline and their proxies participated in a follow-up survey (2001/11). The subjects themselves underwent MMSE again. 136 subject-proxy pairs reported any complaints of memory-related problem and evaluated higher-level functional capacity (TMIG-IC, Tokyo Metropolitan Institute of Gerontology Index of Competence). We established criteria at follow-up survey as follows: control (n = 29), MMSE scores > 1 SD below age-specific means and CDR (Clinical Dementia Rating) = 0: mild cognitive decline (MCD) (n = 54), 21 < or = MMSE scores < 1 SD below age-specific means or CDR = 0.5); and severe cognitive decline (SCD) (n = 53), MMSE scores 20 < or = CDR > 0.5. SCD subjects significantly overestimated total and Instrumental Self-Maintenance scores in TMIG-IC more than control or SCD subjects. Multiple logistic regression analyses indicated that complaints of memory by the proxy, response by spouse, and higher levels of education were extracted as significantly independent variables affecting overestimation for functional capacity. On the other hand, aging affected underestimation.