Immunization of transgenic mouse models of Alzheimer disease using amyloid-beta peptide (Abeta) reduces both the Alzheimer disease-like neuropathology and the spatial memory impairments of these mice. However, a therapeutic trial of immunization with Abeta42 in humans was discontinued because a few patients developed significant meningo-encephalitic cellular inflammatory reactions. Here we show that beneficial effects in mice arise from antibodies selectively directed against residues 4-10 of Abeta42, and that these antibodies inhibit both Abeta fibrillogenesis and cytotoxicity without eliciting an inflammatory response. These findings provide the basis for improved immunization antigens as well as attempts to design small-molecule mimics as alternative therapies.