Handgrip was measured, using a modified sphygmomanometer cuff, in a random sample of 487 older people in Edinburgh in 1968–9. The prevalence of dementia was determined by full psychiatric examination. The error score in the Isaacs-Walkey test was also recorded. Five years later, 261 of the original sample were re-examined. In both examinations treated cross-sectionally there was a significant decrease in grip strength as age increased. However, the mean of the five-year change in individuals did not differ significantly from zero and was not related to age. Handgrip was not significantly related to dementia either at the first examination or as a predictor of new dementia at the five-year examination. Mean handgrip was significantly less at the first examination in those who subsequently died, compared with five-year survivors.