The prevailing amyloid hypothesis for Alzheimer's disease (AD) holds that amyloid beta-protein (Abeta) causes neuronal degeneration by forming neurotoxic fibrillar structures. Yet, many aspects of AD pathology and symptoms are not well explained by this hypothesis. Here, we present evidence that neurotoxicity of soluble oligomeric Abeta closely corresponds to the selective neurodegeneration so distinctly manifest in AD. Selectivity was first observed in vitro, where only the human central nervous system neuronal cells were susceptible to soluble oligomeric Abeta. Furthermore, in mouse cerebral slice treated with soluble oligomeric Abeta, selective regiospecific toxicity was evident in the hippocampal CA1, a division important for memory, but not in the CA3 subfield. The fibrillar Abeta, however, killed neurons in all regions of the cerebral slice cultures and also in cerebellar slices. Remarkably, even at the highest soluble oligomeric Abeta concentrations, cerebellar neurons were completely spared, consistent with one of the hallmark features of AD pathology. Our observation of the selective neurodegeneration of soluble oligomeric Abeta to neurons involved in cognitive function may provide a new opportunity for the development of an effective AD therapy as well as elucidating the pathological mechanism of AD.