Alzheimer’s Disease (AD) impairs not only cognitive abilities, such as memory, but also sensory functions, such as vision. The current project investigates memory-vision interactions in the early stage of AD known as amnestic Mild Cognitive Impairment (aMCI), as well as in healthy aging. We focus on memory for visual stimuli, which declines substantially in AD, predicting dementia more than a decade before diagnosis. We propose 3 functional MRI (fRMI) studies, each including 3 groups of participants: aMCI, older adults, and younger adults. Whereas most fMRI studies on memory and aMCI investigated memory processes, we focus on the quality of the information used by these processes or memory representations. To our knowledge, no fMRI study has examined the quality of memory representations in MCI. Our first aim is to investigate the effects of healthy aging and aMCI on visual vs. semantic features of memory representations. There is evidence that healthy aging impairs visual representations in posterior occipito-temporal cortex (OTC) and medial temporal lobe (MTL) regions, but we have preliminary that it spares semantic representations in anterior OTC/MTL. In contrast, aMCIs are impaired in semantic processing and anterior temporal function. We test the hypothesis that visual representations in posterior OTC/MTL are impaired in both aMCI and healthy older adults, whereas semantic representations in anterior OTC/MTL are impaired in aMCI but not in healthy older adults (Hypothesis 1). This dissociation between healthy and pathological aging has important implications for the early diagnosis of AD. Our second aim is to examine the semantic support effect on memory representations in healthy older adults and aMCIs. The semantic support effect refers to the reliable finding that conditions that allow or promote the use of preexistent semantic knowledge tend to attenuate age-related cognitive deficits, including memory deficits. The semantic support effect is weaker in aMCIs, consistent with their semantic processing deficits. Each of the 3 studies investigates a different form of the semantic support effect (levels of processing, schema congruency, and conceptual attention). We test the hypothesis that the semantic support effect enhances the quality of visual memory representations and subsequent visual memory to a greater extent in healthy older adults than in aMCIs (Hypothesis 2). This finding has implications for the use of the semantic support in memory training in these populations. In sum, the proposed studies will systematically examine the neural mechanisms of visual memory deficits in MCI and healthy aging. They will use novel representational similarity analyses that have not been used yet to investigate memory representations in aMCI. The expected results will contribute to methods for diagnosing AD early in the disease and to rehabilitating memory in aMCI and healthy aging, and are hence, highly significant.