Different schemes for treating the electrostatic interactions in molecular dynamics simulations are investigated : charge-group truncation with or without reaction-field correction, atomic truncation with or without reaction-field correction, and Ewald summation. When a reaction-field correction is applied, the influence of the size of the radius selected for the spherical boundary to the continuum is also considered. The different schemes are applied to simple point charge water simulations, and simulated energetic, transport, structural, and dielectric properties are compared. It is concluded that (i) the inclusion of a reaction-field correction in a charge-group truncation scheme induces significant changes in different types of properties, and that a number of properties are not identical to those observed using the Ewald scheme, (ii) when the reaction-field correction is included in an atomic truncation scheme instead, the agreement with the Ewald results is in general improved, and (iii) the increase (decrease) of the radius of the boundary to the continuum by 0.1 (nm) with respect to the cutoff radius induces in both cases a degradation of the simulated properties. Special attention is paid to the calculation of the dielectric permittivity from the simulations. Due to the finite size of the statistical ensembles considered, this property is not assumed to be isotropic, and the degree of anisotropy is used instead as a test for convergence. Since the incorporation of the reaction-field correction into an atomic truncation scheme leads, when systems of high dielectric permittivity are considered, to electrostatic interactions which implicitly contain a (physically reasonable) shifting function and properties which are comparable to those obtained using the Ewald method, this scheme is a clear improvement over a charge-group-based truncation when a reaction-field correction is used in molecular dynamics simulations of noncharged systems.