OBJECTIVES: To examine whether well-known predictors of mortality change their predictive power over time, being reduced or even reversed in the old-old. DESIGN: A multidimensional survey of the Cross-Sectional and Longitudinal Aging Study conducted from 1989 to 1992 with follow-up of mortality after 6, 8, 10, and 12 years since 1989. SETTING: Israel. PARTICIPANTS: Participants (N=1,369) were drawn from a national sample of the Jewish Israeli population aged 75 to 94. MEASUREMENTS: Data included sociodemographic factors and measures of health, physical condition, cognitive performance, and depression. RESULTS: The results showed that age, sex, disability, self-rated health, and marital status predicted mortality and that their predictive power changed over 9 years. CONCLUSION: In the old-old, predictors of mortality changed over time, and their predictive effect eventually diminished. The predictors found to be most significant (age, sex, disability, and self-rated health) support the common cause theory.