To test whether cognitive ability predicts survival from age 79 to 89 years data were collected from 543 (230 male) participants who entered the study at a mean age of 79.1 years. Most had taken the Moray House Test of general intelligence (MHT) when aged 11 and 79 years from which, in addition to intelligence measures at these two time points, relative lifetime cognitive change was calculated. Cognitive domain measures at age 79 included: vocabulary, nonverbal reasoning, verbal declarative memory, and executive functioning. A fluid-type general intelligence component ("gf") was extracted from reasoning, memory, and executive functioning. Socio-demographic, health behaviour, and health status measures were included as covariates in Cox's proportional hazard regression models of the cognition-mortality associations. Having attained the age of 79, "gf," reasoning and memory measured in later-life, and the relative change in cognitive ability (11 to 79 years) are important in predicting mortality. Health and socio-demographic status are possible mechanisms through which cognitive ability predicts mortality in old age. (Contains 2 tables and 1 figure.)