Broadening parameters for three microwave lines of water at 22.2, 183.3, and 380.2 GHz, in a bath of helium atoms, are calculated using accurate molecular scattering 5 matrices obtained from two theoretical potentials presented by Palma et al., J. Chem. Phys. 89, 1401 (1988). For the 22 GHz line results are in substantial agreement with values presented in that work, indicating the accuracy of approximate methods used there. The present work improves the potential energy surfaces, computed from perturbation theory (MP4) and variational interacting correlated fragments (ICF1) wave funcitons, by correcting them for basis set superposition error (BSSE), and recomputes the line broadening using a different procedure for fitting computed energy points. In addition, the entire set of calculations are repeated with a quite different basis set for orbital expansion to establish the reliability of the potential energy surface. We show that adjustments for superposition error are essential, and that broadening cross sections computed from the new surfaces are changed 10%-30% from the old, significantly improving agreement with experiment. The MP4 BSSE adjusted surface appears to be the most accurate, giving room temperature broadenings of 8.9, 11.8, and 10.0 Å2 compared with experimental determinations of 12.2 ± 1.2, 11.9, and 11.2 Å2 for the 22, 183, and 380 GHz lines, respectively. Thus, computed line to line variation is larger than observed. The ICF1 BSSE adjusted results for pressure broadening cross section parallel those from the MP4 BSSE calculations but are about 10% smaller. We believe our computed results are stable with respect to basis set for orbital expansion and that the scattering calculations are accurate. Any theoretical inadequacy has been pinpointed to too few points on the potential energy surface resulting in an inadequate description of the angle dependence. It is not clear whether the present discrepancy between computation and experiment stems from this or from errors in the experimental values, although we show some indication that additional information on the surface might decrease the computed broadenings, worsening agreement with experiment. © 1991 American Institute of Physics.