Several recent reviews have suggested that cognitive rehabilitation may hold promise in the treatment of memory deficits experienced by patients with mild cognitive impairment. In contrast to the previous reviews that mainly focused on outcome, the current review examines key methodological challenges that are critical for designing and interpreting research studies and translating results into clinical practice. Using methodological details from 36 studies, we first examine diagnostic variability and how the use of cutoffs may bias samples toward more severely impaired patients. Second, the strengths and limitations of several common rehabilitative techniques are discussed. Half of the reviewed studies used a multi-technique approach that precludes the causal attribution between any specific technique and subsequent improvement. Third, there is a clear need to examine the dose-response relationship since this information was strikingly absent from most studies. Fourth, outcome measures varied widely and frequently depended on neuropsychological tests with little theoretical justification or ecological relevance. Fifth, we discuss how the variability in each of these other four areas complicates efforts to examine training generalization. Overall, future studies should place greater emphasis on ecologically relevant treatment approaches and outcome measures and we propose a hierarchical model that may aid in this pursuit. (JINS, 2014, 20, 135-151)