Aim The longitudinal relationship between cognitive abilities and subsequent death was investigated among community‐dwelling older Japanese adults. Methods Participants (n = 1060; age range 60–79 years) comprised the first‐wave participants of the National Institute for Longevity Sciences‐Longitudinal Study of Aging. Participants’ cognitive abilities were measured at baseline using the Japanese Wechsler Adult Intelligence Scale‐Revised Short Form, which includes the following tests: Information (general knowledge), Similarities (logical abstract thinking), Picture Completion (visual perception and long‐term visual memory) and Digit Symbol (information processing speed). By each cognitive test score, participants were classified into three groups: the high‐level group (≥ the mean + 1SD), the low–level group (≤ the mean – 1SD) and the middle‐level group. Data on death and moving during the subsequent 15 years were collected and analyzed using the multiple Cox proportional hazard model adjusted for physical and psychosocial covariates. Results During the follow‐up period, 308 participants (29.06%) had died and 93 participants (8.77%) had moved. In the Similarities test, adjusted hazard ratios (HR) of the low‐level group to the high‐level group were significant (HR 1.49, 95% CI 1.02–2.17, P = 0.038). Furthermore, in the Digit symbol test, the adjusted HR of the low‐level group to the high‐level group was significant (HR 1.62, 95% CI 1.03–2.58, P = 0.038). Significant adjusted HR were not observed for the Information or Picture Completion tests. Conclusions It is suggested that a lower level of logical abstract thinking and slower information processing speed are associated with shorter survival among older Japanese adults. Geriatr Gerontol Int 2016; ••: ••–••.