Estrogens are potent and efficacious neuroprotectants both in vitro and in vivo in a variety of models of neurotoxicity. We determined the structural requirements for neuroprotection in an in vitro assay using a panel of > 70 novel estratrienes, synthesized to reduce or eliminate estrogen receptor (ER) binding. We observed that neuroprotection could be enhanced by as much as 200-fold through modifications that positioned a large bulky group at the C2 or C4 position of the phenolic A ring of the estratriene. Further, substitutions on the B, C or D rings either reduced or did not markedly change neuroprotection. Collectively, there was a negative correlation between binding to ERs and neuroprotection with the more potent compounds showing no ER binding. In an in vivo model for neuroprotection, transient cerebral ischemia, efficacious compounds were active in protection of brain tissue from this pro-oxidant insult. We demonstrated that these non-feminizing estrogens engage in a redox cycle with glutathione, using the hexose monophosphate shunt to apply cytosolic reducing potential to cellular membranes. Together, these results demonstrate that non-feminizing estrogens are neuroprotective and protect brain from the induction of ischemic- and Alzheimer's disease (AD)-like neuropathology in an animal model. These features of non-feminizing estrogens make them attractive compounds for assessment of efficacy in AD and stroke, as they are not expected to show the side effects of chronic estrogen therapy that are mediated by ER actions in the liver, uterus and breast.