The dosimetry of 106Ru/106Rh beta-emitting eye applicators is still inadequate. Manufacturer's specifications of absolute dose rates are loaded with a +/- 30% error, and the relative distribution of activity on the surface is measured on a few points only, using a 3-mm plastic scintillation probe. While reducing the absolute error of dose rate measurements to +/- 15% by a method published earlier, we now present a phantom using small thermoluminescent dosimetry crystals for refined assessment of its distribution over the surface. Evaluation of two applicators revealed a 50% increase in activity in the midperiphery in one and a steep falloff of activity 1 mm within the margin of both specimens. The method and findings are demonstrated in detail and compared to those reported by other authors.