This paper presents the performance results of bioretention pilot experiments conducted in Bryan, Texas, a hot and semiarid area. Five steel boxes were constructed. Each of four boxes was filled with an underdrain pipe, gravel, pea gravel, and soilcompost media and planted with a specific type of vegetation, including shrubs, grass species specified for highways in Texas, native grasses, and Bermuda grass (Cynodon dactylon). The fifth box, the control, had no vegetation planting but contained the other materials. Performance testing was conducted 14 months after construction. Synthetic runoff that mimicked highway storm water runoff characteristics was placed into the pilot boxes, and the concentrations of various pollutants in influent and effluent samples were measured. The results suggested that the pilot boxes effectively removed zinc, lead, total suspended solids, and ammonianitrogen, but exported copper, nitratenitrogen (NO3-N), total nitrogen (TN), and total phosphorus. The leaching out of the boxes by pollutants could be attributed to the use of compost incorporated in the soilcompost media. Vegetation plays an important role in NO3-N and TN removals through root uptake and denitrification processes in the root zone. This paper, based on the results of the study, discusses implications for designing bioretention facilities in a hot, semiarid region.