The overall goal of this project is to obtain an accurate statistical map of memory activation in the hippocampus (and associated structures in the limbic system such as cingulate cortex) in older subjects, age 60 to 75. The successful completion of our objectives may lead to an fMRI-based diagnostic test for memory impairment to characterize abnormal memory function in people at risk for Alzheimer's disease (AD). AD is characterized by memory loss with pathological changes especially in the temporal lobes and hippocampus. The present application seeks to expand on previous studies to produce robust maps of hippocampal and cingulate cortex activation. Using local Canonical Correlation Analysis (CCA), a powerful multivariate statistical method that uses locally adaptive spatial filter kernels, it is possible to detect activation in the hippocampus more reliably than with univariate methods. The present proposal represents a first step toward characterizing memory function, especially for people at risk for developing AD and other forms of dementia. It is divided into two major phases and will be applied to two subject groups (cognitively normal subjects and subjects with mild amnestic cognitive impairment (MCI)). In phase I (year 1, method development), two memory paradigms will be optimized. The goal in this phase is to establish a consistent fMRI methodology (paradigm and analysis method) providing reliable results of hippocampal activation as measured by adaptive CCA (an improved version of CCA). Data collection and fMRI analysis will be carried out at STesIa for 10 healthy control subjects. Activation in the cingulate cortex will also be measured. In phase II (year 2, clinical application) 40 subjects will be scanned: 20 normal subjects and 20 with mild cognitive impairment. Activations of the Hippocampus and other areas of the limbic system (such as cingulate cortex) will be measured. The performance of adaptive CCA will be compared with a standard univariate analysis using SPM2. We will focus initially on group analysis (i.e., investigating differences in memory activation between each group), but in the later phase we will explore individual markers and provide individual confidence intervals to characterize memory deficit, which might predict future development of dementia. ? ? ?