Prescribed fire is increasing being used to reduce hazardous fuel accumulations on National Forest System and other public lands. Those same lands are often home to rare species, for which little or no information on fire response exists. Land managers are stuck with a dilemma – carry out a prescribed fire in rare species habitat with unknown positive or negative effects, or risk having the habitat burn up in a wildfire, again with unknown impacts to species that may already be nearing extinction. Most managers and biological specialists find unacceptable the risk of applying fire to a significant portion of the range of a rare species when nothing is known about how the species will respond. Studies are urgently needed that will yield fire response information without jeopardizing a large part of the remaining population of species in peril. Arabis johnstonii (Johnston’s rock cress), a herbaceous perennial plant found on the San Bernardino National Forest (SBNF), CA, is considered rare by the California Native Plant Society and has been designated a Regional Forester’s Sensitive species by the Pacific Southwest Region, USDA Forest Service. Much of its habitat lies within an area that the SBNF plans to treat with prescribed fire to reduce fuel loading for community protection, but nothing is known about its response to fire. Before subjecting entire populations of this rare plant to prescribed fire of unknown consequence, SBNF staff worked with us to design a small-scale test of fire effects on individual patches of A. johnstonii. Study implementation required developing a safe method to conduct small but relatively hot burns in a wildland area suffering from four years of drought. We constructed a circular “fire cage” (1.1 m diameter x 0.62 m tall) using 8x8 mesh stainless steel woven wire cloth as a firewall. Trials conducted in the Riverside Fire Lab burn building using excelsior, dried chaparral branches packed to 0.5 m depth, and 7 mph wind speeds were successful in containing all potential fire brands and sparks. The fire cage is large enough so that a 10 cm burned buffer zone can surround each treated A. johnstonii patch. Test burns are planned for August and December 2003 to examine the impact of different burning seasons on the plants and seed bank. At each of three population locations, we established 12 (~1 m diameter) plots with a minimum of five A. johnstonii plants in each plot. Fire treatments were randomly assigned to 6 plots per site (3 each for August and December) and were paired with control plots. For successful combustion on plots, excelsior or on-site fuels may be added if natural fuel is too sparse to carry fire. Prior to ignition, a 2-cm wide scratch line to inorganic soil will be dug around the perimeter of the fire cage. Fire shelters will be placed around the fire cage to serve as an additional 2 m wide fuel break. Fuels will be ignited with a butane torch. Weather, fuel conditions, and fire behavior variables will be documented during the burns. Arabis johnstonii survival, growth and recruitment will be measured twice a year for up to 3 years after the test fires, beginning in spring 2004. Data from the August 2003 experimental burns will be presented and should illustrate how this technique performs under field conditions. Resource specialists may choose to use this method to investigate the fire tolerance of other plant species of concern, in order to help managers anticipate the impacts of prescribed fire and plan its implementation in a way that will minimize negative impacts to rare species.