Experimental data on the slow-neutron ( < 0.1 eV) scattering of room-temperature water have been obtained using the velocity selector at the Materials Testing Reactor. The data agree with the results of Haywood obtained using a slow-neutron velocity selector, with the results of Kottwitz and Leonard obtained using a triple-axis spectrometer, and with the results of Kirouac obtained using a linear accelerator plus phased chopper. There are marked differences from the results of Sakamoto and co-workers, who used cold neutrons and a rotating crystal spectrometer. An extension of the Nelkin model has been devised that gives calculated results in good agreement with the MTR data, whereas the Nelkin model generally gives poor agreement. The new model treats water as a mixture of 10% of free molecules and 45% each of two aggregates in which the effective masses for H scattering are 75 and 150. Each aggregate has a set of vibrational modes with energies distributed in the range 0 < E ≤ 0.125 eV, and each H2O molecule exhibits the internal modes with characteristic energies of 0.2, 0.45, and 0.46 eV. The amplitudes of the low-energy vibrations are selected so the calculated scattering agrees with the Materials Testing Reactor data. Total cross sections calculated on the new model and the Nelkin model agree quite well with the data. The new model gives values of the average cosine of the scattering angle that are closer to the data than those calculated by the Nelkin model, but the calculated results are always higher than the data. At very low neutron energies, the present Nelkin model calculations give total cross sections that are lower than those reported by Koppel and Young using essentially the same model. The reason for this is not known.