Accurately predicting the development of probable Alzheimer's disease (AD) early in the course of the disease would have major implications. Hence, the pre-dementia stage of AD has become a major topic of current research. Amnestic 'mild cognitive impairment' (MCI) has recently emerged as the most convenient avenue to address this Issue, since most of MCI patients will progress to AD, at a rate of 10% to 15% per Year, suggesting that MCI represents the clinical manifestation of incipient AD (Petersen et al., 2001). However, all MCI patients would probably not convert to AD, at least in the near future, so that the Issue of prospective longitudinal studies is to detect specific indices of rapid conversion. In this paper, we focus on longitudinal studies using either neuropsychology, or morphological or functional neuroimaging to find predictive markers allowing to distinguish those MCI patients that rapidly convert to AD from those that do not develop the disease during the follow-up period. Whereas functional neuroimaging, and more specifically FDG-PET, seems particularly accurate to predict AD, the combination of multiple approaches is likely to be a promising avenue.