BACKGROUND: We evaluated the effects of cognitive training on self-rated health at 1, 2, 3, and 5 years post-baseline. METHODS: In the ACTIVE (Advanced Cognitive Training for Independent and Vital Elderly) randomized controlled trial, 2,802 older adults (>or=65 years) were randomly assigned to memory, reasoning, speed of processing, or no-contact control intervention groups. Complete data were available for 1,804 (64%) of the 2,802 participants at five years. A propensity score model was adjusted for attrition bias. The self-rated health question was coded using the Diehr et al. (2001) transformation (E = 95/VG = 90/G = 80/F = 30/P = 15), and analyzed with change-score regression models. RESULTS: The speed of processing (vs. no-contact control) group had statistically significant improvements (or protective effects) on changes in self-rated health at the 2, 3 and 5 year follow-ups. The 5-year improvement was 2.8 points (p = 0.03). No significant differences were observed in the memory or reasoning groups at any time. CONCLUSION: The speed of processing intervention significantly protected self-rated health in ACTIVE, with the average benefit equivalent to half the difference between excellent vs. very good health.