Successful recovery of the federally threatened southeastern beach mouse (Peromyscus polionotus niveiventris) depends in part on an understanding of their habitat requirements. I studied habitat use by beach mice at Cape Canaveral Air Force Station, Florida from March 2005 until March 2006. I livetrapped six grids, three on coastal dunes and three within scrub located inland from the coast. On each grid and trap station, I quantified the extent of bare ground, woody vegetation, non-woody vegetation, height of vegetation, and percentage of coarse sand in the surface soil. I assessed trap success relative to these habitat variables using linear and multiple regression, correlation, and ordination. Significantly higher numbers of mice were captured in the scrub habitat relative to the coastal habitat. Linear regression of trap success against the habitat variables did not reveal any significant relationships at the level of grids. A non-metric multidimensional scaling model was designed to capture the vegetation heterogeneity at the trapping sites and clarify the results. This methodology identified a predominantly dune and predominately scrub cluster of trap sites. A bubble plot showed higher densities of beach mice using the scrub habitat types. These results suggest beach mice are selecting for those habitat variables defined by the ordination: higher vegetation height, more woody vegetation types, less bare ground, and less heterogeneity.