Performance usually improves when cognitive tests are retaken, and the magnitude of improvement may predict Alzheimer’s disease pathology in older adults (Duff et al., 2014). We tested 172 healthy older adults (mean age 72 years) on consecutive days to measure overnight retest learning, then retested them at 6, 18, and 30 months to measure cognitive trajectory. Subjects with >1 SD below-average overnight learning gain had 4x the odds of being in the lowest 30-month cognitive trajectory group.