We present benchmark calculations of vertical electron detachment energies (VDEs) for various conformers of (H2O)n-, using both wave function and density functional methods, in sequences of increasingly diffuse Gaussian basis sets. For small clusters (n < or = 6), a systematic examination of VDE convergence reveals that it is possible to converge this quantity to within approximately 0.01 eV of the complete-basis limit, using a highly diffuse but otherwise economical Pople-style basis set of double-zeta quality, with 28 atom-centered basis functions per water molecule. Floating-center basis functions can be useful but are not required to obtain accurate VDEs. Second-order Møller-Plesset perturbation (MP2) theory suffices to obtain VDEs that are within 0.05 eV of the results from both experiment and coupled-cluster theory, and which always err toward underbinding the extra electron. In contrast to these consistent predictions, VDEs calculated using density functional theory (DFT) vary widely, according to the fraction of Hartree-Fock exchange in a given functional. Common functionals such as BLYP and B3LYP overestimate the VDE by 0.2-0.5 eV, whereas a variant of Becke's "half and half" functional is much closer to coupled-cluster predictions. Exploratory calculations for (H2O)20- and (H2O)24- cast considerable doubt on earlier calculations that were used to assign the photoelectron spectra of these species to particular cluster isomers.