Over the past several years, experiments with synthetic amyloid-beta peptide (Abeta) and animal models have strongly suggested that pathogenesis of Alzheimer's disease (AD) involves soluble assemblies of Abeta peptides (Trends Neurosci. 24 (2001) 219). These soluble neurotoxins (known as ADDLs and protofibrils) seem likely to account for the imperfect correlation between insoluble fibrillar amyloid deposits and AD progression. Recent experiments have detected the presence of ADDLs in AD-afflicted brain tissue and in transgenic-mice models of AD. The presence of high affinity ADDL binding proteins in hippocampus and frontal cortex but not cerebellum parallels the regional specificity of AD pathology and suggests involvement of a toxin receptor-mediated mechanism. The properties of ADDLs and their presence in AD-afflicted brain are consistent with their putative role even in the earliest stages of AD, including forms of mild cognitive impairment.