Generation of reactive oxygen species (ROS) causes cellular oxidative damage and has been implicated in the etiology of Alzheimer's disease (AD). In contrast, multiple lines of evidence indicate that ROS can normally modulate long-term potentiation (LTP), a cellular model for memory formation. We recently showed that decreasing the level of superoxide through the overexpression of mitochondrial superoxide dismutase (SOD-2) prevents memory deficits in the Tg2576 mouse model of AD. In the current study, we explored whether AD-related LTP impairments could be prevented when ROS generation from mitochondria was diminished either pharmacologically or via genetic manipulation. In wild-type hippocampal slices treated with exogenous amyloid β peptide (Aβ1-42) and in slices from APP/PS1 mutant mice that model AD, LTP was impaired. The LTP impairments were prevented by MitoQ, a mitochondria-targeted antioxidant, and EUK134, an SOD and catalase mimetic. In contrast, inhibition of NADPH oxidase either by diphenyliodonium (DPI) or by genetically deleting gp91(phox), the key enzymatic component of NADPH oxidase, had no effect on Aβ-induced LTP blockade. Moreover, live staining with MitoSOX Red, a mitochondrial superoxide indicator, combined with confocal microscopy, revealed that Aβ-induced superoxide production could be blunted by MitoQ, but not DPI, in agreement with our electrophysiological findings. Finally, in transgenic mice overexpressing SOD-2, Aβ-induced LTP impairments and superoxide generation were prevented. Our data suggest a causal relationship between mitochondrial ROS imbalance and Aβ-induced impairments in hippocampal synaptic plasticity.