Background Numerous stroke studies have controversially shown estrogens to be either neuroprotective or neurodamaging. The discordant results observed in rat brain ischemia models may be a consequence of discrepancies in estrogen administration modes resulting in plasma concentration profiles far from those intended. To test this hypothesis we reproduced in detail and extended an earlier study from our lab using a different mode of 17β-estradiol administration; home-made silastic capsules instead of commercial slow-release 17β-estradiol pellets. Four groups of female rats (n = 12) were ovariectomized and administered 17β-estradiol or placebo via silastic capsules. All animals underwent MCAo fourteen days after ovariectomy and were sacrificed three days later. Results In contrast to our earlier results using the commercial pellets, the group receiving 17β-estradiol during the entire experiment had significantly smaller lesions than the group receiving placebo (mean ± SEM: 3.85 ± 0.70% versus 7.15 ± 0.27% of total slice area, respectively; p = 0.015). No significant neuroprotection was found when the 17β-estradiol was administered only during the two weeks before or the three days immediately after MCAo. Conclusions The results indicate that different estrogen treatment regimens result in diametrically different effects on cerebral ischemia. Thus the effects of estrogens on ischemic damage seem to be concentration-related, with a biphasic, or even more complex, dose-response relation. These findings have implications for the design of animal experiments and also have a bearing on the estrogen doses used for peri-menopausal hormone replacement therapy.