The aim of the current study was to test the accuracy of practice effects, that is, improvement in test performance due to repeated neuropsychological evaluations, in identifying patients with amnestic mild cognitive impairment (a-MCI) at greater risk of conversion to Alzheimer disease (AD). For this purpose, we conducted a longitudinal study of 54 patients diagnosed with a-MCI at the first assessment and followed-up for 4 years. During this time, 18 patients converted to AD. Baseline and 6- to 12-month follow-up performances on a large set of neuropsychological tests were analyzed to determine their diagnostic ability to predict later conversion to dementia. Results demonstrate that a lack of practice effects on episodic memory tests is an accurate prognostic indicator of late conversion to AD in a-MCI patients. In fact, even though the performance of both groups was substantially comparable at the baseline evaluation, stable a-MCI patients greatly improved their memory performance at retest after 6 to 12 months; instead, scores of converter a-MCI remained stable or decreased passing from baseline to follow-up. Standardized z-change scores on memory tasks, which were computed as a reliable measure of performance change, classified group membership with very good overall accuracy, which was higher than the classification of converter and stable a-MCIs provided by baseline or follow-up scores. We hypothesize that the lack of practice effects on memory tasks mirrors the early involvement of medial temporal lobe areas in converter a-MCI that are fundamental for the consolidation of new memory traces.