There is accumulating evidence that participation in cognitive leisure activities might be related to a reduced risk of cognitive impairment. However, there is no consensus regarding the classification of cognitive leisure activities, and it cannot clearly define each activity and quantitatively evaluate the effect of it. In the present study, we propose a new classification method, and examine the relationship between reclassified cognitive leisure activities and cognitive function. The analysis included 4830 participants of the Chinese Longitudinal Healthy Longevity Survey, who were aged >80 years and free of cognitive impairment at 1998. They were followed up six times. Six cognitive leisure activities were mainly grouped into intrapersonal and interpersonal cognitive leisure activity by categorical principal component analysis. Cox proportional hazards analysis was used to evaluate the relationship between reclassified cognitive leisure activities and the risk of cognitive impairment. After a median follow-up period of 16 years, 1763 participants suffered from cognitive impairment (Mini-Mental State Examination score <24). Reading newspapers or books, taking part in some social activities and sometimes playing cards or mahjong were associated with a decreased risk of cognitive impairment (P < 0.05 for all). In multivariable adjusted models, compared with those with lower intensities, the hazard ratios (95% confident intervals) of cognitive impairment were 0.86 (0.74-0.99) and 0.71 (0.60-0.83) for participants with high intensities of intrapersonal and interpersonal cognitive leisure activity, respectively. A high frequency of participation in intrapersonal and interpersonal cognitive leisure activities showed a lower risk of cognitive impairment among the oldest-old in China. Geriatr Gerontol Int 2019; 19: 1041-1047.