[en] An analysis of experimental data from 21 fast reactor shield configurations containing steel, sodium, and iron were made as part of a study of the upper axial shielding needs of the Clinch River Breeder Reactor. The measured data were analyzed using both one- and two-dimensional discrete ordinates transport codes and several cross section libraries based on ENDF/B-IV data with group structures of 51 and 171 neutron groups. One-dimensional sensitivity studies using the 171 group library and ENDF/B-IV covariance files for sodium and iron data were used to determine the sensitivities of the measured data to multigroup cross sections and to estimate uncertainties in the calculated results. Results indicate that the standard 51-group design cross section library could be expected to predict the measurements to within 30% over 12 decades of attenuation although a few of the deepest penetration configurations showed disagreements as large as a factor of three. The sensitivity results revealed very high sensitivity of the measurements to total cross section minima and cross sections from 5 to 10 MeV in sodium and iron in the deep penetration configurations. As a result, large uncertainties in the calculated results arose from small uncertainties in the cross section data. These results indicate the need for better measurements of the total cross section minima in sodium, especially around 300 keV