Recommendations for clinical trials methods for 'pre-dementia,' 'prodromal,' or early Alzheimer's disease are discussed. Early AD can be considered as subsets of both 'amnestic MCI' and 'probable AD.' In principle, it can be operationalized using recently proposed, new research criteria for AD that specifically does not require impairment in non-memory cognitive function and activities of daily living, and consequently does not require the presence of dementia. The criteria also require patients to show abnormal putative biomarkers but require validation. Trials in early AD should be done when models of drug action and response suggest that the drug in development likely would be effective in early AD and clinical effects could be expected in a relatively short time. Biomarkers should be used as stratification or explanatory variables that may help to explain clinical outcomes from early AD trials rather than as inclusion/exclusion criteria in order to avoid pseudospecificity. Trials should be multicentered, double-blinded, randomized, placebo-controlled, generally with dose-ranging of two doses if indicated. Duration of trials should be based on expected onsets and durations of effects, and generally should be less than one year. Crossover trials should be considered when appropriate. Primary outcomes should specifically assess memory and include repeated assessments. Potential secondary outcomes could include self- and observer-rated health-related quality of life and global impressions of change in lieu of activities of daily living. Onset of dementia should not be an endpoint because many patients would be on the cusp of dementia and dementia onset is influenced by numerous biological and environmental factors. Inferences that can be made from trials results will likely involve the effects of the test drug on memory and self-rated global function. Disease modification is not likely to be inferred except in trials over two years in duration in which a change in a biomarker can be used as an adjunctive assessment. Models and simulations using existing clinical trials databases would be helpful in planning early AD trials.