Thirty-seven Suprasil quartz spheres, each approximately 1 cm in diameter and containing an iodide-iodate actinometric solution, were attached to a metal rack and inserted into a bench-scale UV reactor filled with water. The spheres were located at various distances and heights around a 12.4 W low-pressure Hg lamp housed inside a 3.2 cm-radius quartz sleeve in the middle of an annular batch reactor. UV light exposure at 254 nm was performed with the percent transmittance of the water present in the reactor at either 13% or 100% defined over a 1 cm path length. The spheres were simultaneously exposed to the UV light for a given period of time, after which the solutions were removed from the spheres and the yield of triiodide determined from the increase in absorbance at 352 nm. The resulting fluence rate at each site was then calculated on basis of the yield of triiodide. These results were compared with the predictions of a mathematical model based on the multiple point source summation approximation, including reflection and refraction at the air-quartzwater interface. Initially, the agreement was not satisfactory, especially in regions at an oblique angle to the lamp. The model was modified from a multiple point source model to a multiple cylindrical segment model by incorporating a cosine factor. The agreement between the new model and the experimental data was excellent and these experiments provide a strong validation of the model, even under conditions in which the fluence rate varied by >1000-fold between extreme sites in the reactor.