The pine rocklands of South Florida, characterized by a rich herbaceous flora with many narrowly endemic taxa beneath an overstory of south Florida slash pine (Pinus elliottii var. densa), are found in three areas: the Miami Rock Ridge of southeastern peninsular Florida, the Lower Florida Keys, and slightly elevated portions of the southern Big Cypress National Preserve. Fire is an important element in these ecosystems, since in its absence the pine canopy is likely to be replaced by dense hardwoods, resulting in loss of the characteristic pineland herb flora. Prescribed fire has been used in Florida Keys pine forests since the creation of the National Key Deer Refuge (NKDR), with the primary aim of reducing fuels. Because fire can also be an effective tool in shaping ecological communities, we conducted a 4-year research study which explored a range of fire management options in NKDR. The intent of the study was to provide the Fish and Wildlife Service and other land managers with information regarding when and where to burn in order to perpetuate these unique forests. In 1998 we initiated a burning experiment in a randomized complete block design. Three treatments were to be carried out in a single well-defined block in each of two characteristic understory types during each year from 1998 through 2000. One understory type was characterized by a relatively sparse shrub layer and a well-developed herb layer ('open'), and the second had a dense shrub layer and poorly developed herb layer ('shrubby'). The three burn treatments were: (a) summer burn, (b) winter burn, and (c) no burn, or control. Three 1- ha plots were established in each block, and randomly assigned to the three treatments. Though the first year experimental burns were carried out without incident, constraints posed by external factors, including nationwide and statewide prohibitions on prescribed burning due to wildfires in other regions, delayed the experimental burns and precluded collection of postburn data on one third of the burns. Ultimately we burned only eleven plots, three in winter and eight in summer, over a four-year period from 1998 to 2001. Vegetation was sampled in a stratified, nested design within 18 plots. Trees were sampled in a 1.0-ha plot, shrubs in twenty 50-m2 circular (radius 4 m) subplots within the tree plot, and the herb layer in four circular 1-m2 quadrats (radius 0.57 m) within each subplot. The amount of fuel in the shrub layer was estimated by applying regression models to plant dimensional data, and ground layer fuel was estimated by a harvest method. The effects of Key deer herbivory on regeneration of the understory pine rockland plant community after fire was studied by monitoring inside and outside exclosures established within two of the six blocks. Pine trees constituted more than half (53.3%) of the biomass, but understory fine fuels comprised a surprisingly high proportion of total aboveground biomass. In the three blocks in which paired summer and winter burns were successfully conducted, the summer burns were more intense than the winter burns as judged by our indicators of fire intensity. Because of the differences in fire intensity between seasons, it was not possible to say whether observed differences in vegetation response between summer and winter burns were due to season or to fire intensity. The mortality of South Florida slash pine trees was greater after the summer burn than the winter burn in each block, but other vegetation responses were rarely as consistent. For instance, Metopium showed less recovery after summer burns in two blocks and after the winter burn in the third block. Moreover, there were instances in which alternative growth stages of the same species responded differently. Adult palms succumbed more frequently to summer than winter burns, and mortality of Coccothrinax exceeded that of Thrinax. In contrast, small palms recovered more readily after summer burns than winter burns. High in