Designers of thermal-neutron reactors have always had to adjust microscopic nuclear cross sections to predict neutron multiplication in slightly enriched uranium lattices. It has been surmised that the problem lies in an overestimation of the neutron capture cross section of 238U below 100 eV. We have measured these cross sections by three independent experiments. First, a series of neutron transmission and self-indication measurements were taken on samples of 238U ranging from 10.79 to 11 620 b/atom in inverse thickness. The level parameters were obtained using area analysis and multilevel fits. Next, the capture cross sections deduced from these level parameters were confirmed by direct measurements on both the continuum and discrete line portions of the low-energy gamma-ray spectra. High resolution measurements on the gamma-ray spectra were carried out from 530 to 900 keV over the neutron energy range from near thermal to ∼20 eV. Finally, a further check was made by activating thin samples of 238U with monochromatic neutrons obtained by Bragg scattering. The result is consistent with the capture cross sections obtained by the gamma-ray spectra measurement. Our results reduce, by 25%, the shielded capture integral discrepancy observed in early Bettis Atomic Power Laboratory critical experiments (TRX) with low-235U-enriched uranium rods latticed in water. When they are coupled with refined lattice calculations, much of the long-standing discrepancy is removed.