Well-researched statistical methods are required to guide clinicians in determining the significance of test score changes in serial neuropsychological assessment of older adults. The following six change score methods were examined using five-year test–retest data from the Canadian Study of Health and Aging: the standard deviation method, three reliable change indices (RCIs), and two standardized regression-based methods. Changes in scores on four memory measures were examined in cognitively healthy older adults, and the RCI with a correction for practice/aging effects most accurately classified this normal variability. Diagnostic change (i.e., developing dementia versus remaining cognitive healthy) was also examined in relation to memory test score changes. All change score methods were significantly associated with diagnostic change, though the strength of association varied by measure and method. In contrast to some previous research, RCIs were found to be useful when making diagnostic discriminations in older adults.