Background: The association between decline in handgrip strength from midlife to late life and dementia is unclear. Methods: Japanese community-dwellers without dementia aged 60 to 79 years (ie, individuals in late life; mean age, 68 years) were followed for 24 years (1988-2012) (n = 1,055); 835 of them had participated in a health examination in 1973-1974 (mean age, 53 years), and these earlier data were used for the midlife analysis. Using a Cox proportional hazards model, we estimated the risk conferred by a decline in handgrip strength over a 15-year period (1973-74 to 1988) from midlife to late life on the development of total dementia, Alzheimer's disease (AD), and vascular dementia (VaD) over the late-life follow-up period from 1988 to 2012. Results: During the follow-up, 368 subjects experienced total dementia. The age-and sex-adjusted incidence of total dementia increased significantly with greater decline in handgrip strength (increased or unchanged handgrip strength [≥+0%] 25.1, mildly decreased [−14 to −1%] 28.4, and severely decreased [≤−15%] 38.9 per 1,000 person-years). A greater decline in handgrip strength was significantly associated with higher risk of total dementia after adjusting for potential confounding factors; subjects with severely decreased handgrip strength had 1.51-fold (95% confidence interval, 1.14-1.99, P < 0.01) increased risk of total dementia compared to those with increased or unchanged handgrip strength. Similar significant findings were observed for AD, but not for VaD. Conclusions: Our findings suggest that a greater decline in handgrip strength from midlife to late life is an important indicator for late-life onset of dementia.